CIFAR_exps_learningrates.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "CIFAR_exps.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "Lu83Jf5qHY_l",
        "colab_type": "code",
        "outputId": "5a9ff315-c5e7-4ac7-b01e-58b4887da861",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 63
        }
      },
      "source": [
        "import torchvision\n",
        "import tensorflow as tf"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<p style=\"color: red;\">\n",
              "The default version of TensorFlow in Colab will soon switch to TensorFlow 2.x.<br>\n",
              "We recommend you <a href=\"https://www.tensorflow.org/guide/migrate\" target=\"_blank\">upgrade</a> now \n",
              "or ensure your notebook will continue to use TensorFlow 1.x via the <code>%tensorflow_version 1.x</code> magic:\n",
              "<a href=\"https://colab.research.google.com/notebooks/tensorflow_version.ipynb\" target=\"_blank\">more info</a>.</p>\n"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oYTXiSckMwDa",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#%cd /content/\n",
        "#!/usr/bin/python\n",
        "# essential imports\n",
        "from __future__ import absolute_import\n",
        "from __future__ import division\n",
        "from __future__ import print_function\n",
        "\n",
        "# Ignore warnings\n",
        "import warnings\n",
        "warnings.filterwarnings(\"ignore\")\n",
        "\n",
        "\n",
        "import imageio\n",
        "import cv2\n",
        "import os\n",
        "import sys\n",
        "import random\n",
        "import matplotlib.pyplot as plt\n",
        "%matplotlib inline\n",
        "plt.ion() \n",
        "import numpy as np\n",
        "import timeit\n",
        "import pandas as pd\n",
        "import re\n",
        "import math\n",
        "import copy\n",
        "import pickle\n",
        "#tensorflow 1.15\n",
        "import tensorflow as tf\n",
        "#print(tf.__version__)\n",
        "import torch\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "# default seeding for reproducability\n",
        "def seed_everything(seed):\n",
        "    random.seed(seed)\n",
        "    os.environ['PYTHONHASHSEED'] = str(seed)\n",
        "    np.random.seed(seed)\n",
        "    torch.manual_seed(seed)\n",
        "    torch.cuda.manual_seed(seed)\n",
        "    torch.backends.cudnn.deterministic = True\n",
        "\n",
        "seed_everything(24)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3MXcpFQSOI50",
        "colab_type": "code",
        "outputId": "74185832-e836-4968-fff6-e12fc8820019",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 142
        }
      },
      "source": [
        "# Import torch Libraries\n",
        "import torch\n",
        "from torch import optim, nn\n",
        "import torchvision\n",
        "from torchvision import datasets, models, transforms\n",
        "import torch.nn.functional as F\n",
        "\n",
        "from torch.nn import Linear, ReLU, CrossEntropyLoss, MSELoss, Sequential, Conv2d, MaxPool2d, Module, Softmax, BatchNorm2d, Dropout\n",
        "#from torch.autograd import Variable\n",
        "from torch.optim import Adam, SGD\n",
        "from torch.optim import lr_scheduler\n",
        "#adam sgd combined optimizer\n",
        "!pip install adabound\n",
        "import adabound\n",
        "# for evaluating the model\n",
        "from sklearn.metrics import accuracy_score\n",
        "from tqdm import tqdm\n",
        "\n",
        "import copy\n",
        "\n",
        "#defaults\n",
        "PIXEL_LENGTH_MODIFIED = 128\n",
        "FEATURE_SIZE_MODIFIED = PIXEL_LENGTH_MODIFIED*PIXEL_LENGTH_MODIFIED\n",
        "PIXEL_LENGTH_MNIST = 28\n",
        "FEATURE_SIZE_MNIST = PIXEL_LENGTH_MNIST*PIXEL_LENGTH_MNIST"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting adabound\n",
            "  Downloading https://files.pythonhosted.org/packages/cd/44/0c2c414effb3d9750d780b230dbb67ea48ddc5d9a6d7a9b7e6fcc6bdcff9/adabound-0.0.5-py3-none-any.whl\n",
            "Requirement already satisfied: torch>=0.4.0 in /usr/local/lib/python3.6/dist-packages (from adabound) (1.3.1)\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.6/dist-packages (from torch>=0.4.0->adabound) (1.17.4)\n",
            "Installing collected packages: adabound\n",
            "Successfully installed adabound-0.0.5\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "z3Tb-VZDOU5Q",
        "colab_type": "text"
      },
      "source": [
        "## helpers"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bLScvSjuOWFZ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# training/testing functions\n",
        "def train(epoch, train_loader, model, error, optimizer, batch_size):\n",
        "    model.train()\n",
        "    running_loss = 0.0\n",
        "    #loss_total = 0.0\n",
        "    for batch_idx, (images, labels) in enumerate(train_loader):\n",
        "        #peak at tensor details\n",
        "        # if batch_idx==1:\n",
        "        #   # visualize one of the images in data set\n",
        "        #   plt.imshow(np.squeeze(images[0].numpy()), cmap='Greys')\n",
        "        #   plt.axis(\"off\")\n",
        "        #   plt.title(str(labels[0].numpy()))\n",
        "        #   #plt.savefig('graph.png')\n",
        "        #   plt.show()\n",
        "        #reshape for training\n",
        "        train = images.view(batch_size,3,32,32).to(device=device, dtype=torch.float)\n",
        "        labels = labels.to(device=device, dtype=torch.long)\n",
        "        # Clear gradients\n",
        "        optimizer.zero_grad()\n",
        "        # Forward propagation\n",
        "        outputs = model(train)\n",
        "        # Calculate softmax and cross entropy loss\n",
        "        loss = error(outputs, labels)\n",
        "        #print(loss)\n",
        "        #loss_total += loss.item()\n",
        "\n",
        "        # Calculating gradients\n",
        "        loss.backward()\n",
        "        # Update parameters\n",
        "        optimizer.step()\n",
        "        #print every 100 batches\n",
        "        running_loss += loss.item()\n",
        "        if batch_idx % 100 == 0 and batch_idx != 0:\n",
        "            # plt.imshow(images[0].numpy().reshape(128,128))\n",
        "            # plt.axis('off')#########################################################################################\n",
        "            # plt.show()\n",
        "            #Print Loss\n",
        "            print('Train Epoch: {} [{}/{} ({:.0f}%)]\\tTrain Loss: {:.6f}'.format(\n",
        "                epoch, batch_idx * len(images), len(train_loader.dataset),\n",
        "                100. * batch_idx / len(train_loader), running_loss/100))\n",
        "            running_loss = 0.0\n",
        "    #loss_total /= len(train_loader)\n",
        "    #print('Test set: Test loss: {:.4f}'.format(loss_total))\n",
        "    #return last loss\n",
        "    return loss.item()\n",
        "\n",
        "def test(test_loader, model, error, batch_size):\n",
        "    # Validation\n",
        "    with torch.no_grad():\n",
        "        model.eval()\n",
        "        # Calculate Accuracy         \n",
        "        correct = 0\n",
        "        loss_test = 0.0\n",
        "        # Iterate through test dataset\n",
        "        for batch_idx, (images, labels) in enumerate(test_loader):\n",
        "            # if batch_idx==1:\n",
        "            #   # visualize one of the images in data set\n",
        "            #   plt.imshow(np.squeeze(images[0].numpy()), cmap='Greys')\n",
        "            #   plt.axis(\"off\")\n",
        "            #   plt.title(str(labels[0].numpy()))\n",
        "            #   #plt.savefig('graph.png')\n",
        "            #   plt.show()\n",
        "\n",
        "            test = images.view(batch_size,3,32,32).to(device=device, dtype=torch.float)\n",
        "            labels = labels.to(device=device, dtype=torch.long)\n",
        "            # Forward propagation\n",
        "            outputs = model(test)\n",
        "            \n",
        "\n",
        "            # sum up batch loss\n",
        "            loss_test += error(outputs, labels).item()\n",
        "            # get the index of the max log-probability\n",
        "            predicted = outputs.max(1, keepdim=True)[1]\n",
        "            # if batch_idx==1:\n",
        "            #   print(predicted[0].item())\n",
        "            correct += predicted.eq(labels.view_as(predicted)).sum().item()\n",
        "        loss_test /= len(test_loader)\n",
        "        accuracy = 100. * correct / len(test_loader.dataset)\n",
        "        print('\\nTest set: Test loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\\n'\n",
        "              .format(loss_test, correct, len(test_loader.dataset), accuracy))\n",
        "        #return loss and accuracy\n",
        "        return loss_test, accuracy\n",
        "        "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dK6XR7HuS4xp",
        "colab_type": "code",
        "outputId": "9d25c392-e41a-4241-a5c2-85886c6f9231",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 107
        }
      },
      "source": [
        "batch_size = 50\n",
        "workers = os.cpu_count()\n",
        "\n",
        "transform = transforms.Compose(\n",
        "    [transforms.ToTensor(),\n",
        "     transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])\n",
        "\n",
        "lengths = [5000, 5000]\n",
        "\n",
        "fullset_train = torchvision.datasets.CIFAR10(root='./data', download=True, train=True,\n",
        "                                       transform=transform)\n",
        "\n",
        "fullset_test = torchvision.datasets.CIFAR10(root='./data', download=True, train=False,\n",
        "                                       transform=transform)\n",
        "\n",
        "testset, finaltestset = torch.utils.data.random_split(fullset_test, lengths)\n",
        "\n",
        "train_loader = torch.utils.data.DataLoader(fullset_train, batch_size=batch_size,\n",
        "                                          shuffle=True, num_workers=workers)\n",
        "\n",
        "test_loader = torch.utils.data.DataLoader(testset, batch_size=batch_size,\n",
        "                                         shuffle=False, num_workers=workers)\n",
        "\n",
        "final_test_loader = torch.utils.data.DataLoader(finaltestset, batch_size=batch_size,\n",
        "                                         shuffle=False, num_workers=workers)\n",
        "\n",
        "#check for gpu/cpu\n",
        "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "  0%|          | 0/170498071 [00:00<?, ?it/s]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Downloading https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz to ./data/cifar-10-python.tar.gz\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "170500096it [00:01, 95080851.94it/s]                               \n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Extracting ./data/cifar-10-python.tar.gz to ./data\n",
            "Files already downloaded and verified\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "143Pa1FYjLfS",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eXhP2GvLy-pw",
        "colab_type": "code",
        "outputId": "2b12ac38-5f24-4263-8acf-e45955c023b4",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 175
        }
      },
      "source": [
        "fullset_test"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Dataset CIFAR10\n",
              "    Number of datapoints: 10000\n",
              "    Root location: ./data\n",
              "    Split: Test\n",
              "    StandardTransform\n",
              "Transform: Compose(\n",
              "               ToTensor()\n",
              "               Normalize(mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5))\n",
              "           )"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "n_y6fbLsVBbd",
        "colab_type": "code",
        "outputId": "8f4b0f23-1130-4871-a537-97240222aba6",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "train_loader.dataset.data.shape"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(50000, 32, 32, 3)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6uUybJbjN6W_",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gTFC4-mQZMse",
        "colab_type": "text"
      },
      "source": [
        "## squeeze dat net"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "x2C4MQMcQbAA",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.nn.init as init\n",
        "import torch.utils.model_zoo as model_zoo\n",
        "\n",
        "__all__ = ['SqueezeNet', 'squeezenet1_0', 'squeezenet1_1']\n",
        "\n",
        "model_urls = {\n",
        "    'squeezenet1_0': 'https://download.pytorch.org/models/squeezenet1_0-a815701f.pth',\n",
        "    'squeezenet1_1': 'https://download.pytorch.org/models/squeezenet1_1-f364aa15.pth',\n",
        "}\n",
        "\n",
        "\n",
        "class Fire(nn.Module):\n",
        "\n",
        "    def __init__(self, inplanes, squeeze_planes,\n",
        "                 expand1x1_planes, expand3x3_planes):\n",
        "        super(Fire, self).__init__()\n",
        "        self.inplanes = inplanes\n",
        "        self.squeeze = nn.Conv2d(inplanes, squeeze_planes, kernel_size=1)\n",
        "        self.squeeze_activation = nn.ReLU(inplace=True)\n",
        "       \n",
        "        \n",
        "        self.expand1x1 = nn.Conv2d(squeeze_planes, expand1x1_planes,\n",
        "                                   kernel_size=1)\n",
        "        self.expand1x1_activation = nn.ReLU(inplace=True)\n",
        "        self.expand3x3 = nn.Conv2d(squeeze_planes, expand3x3_planes,\n",
        "                                   kernel_size=3, padding=1)\n",
        "        self.expand3x3_activation = nn.ReLU(inplace=True)\n",
        "        \n",
        "        \n",
        "    def forward(self, x):\n",
        "        x = self.squeeze_activation(self.squeeze(x))\n",
        "        return torch.cat([\n",
        "            self.expand1x1_activation(self.expand1x1(x)),\n",
        "            self.expand3x3_activation(self.expand3x3(x))\n",
        "        ], 1)\n",
        "\n",
        "\n",
        "class SqueezeNet(nn.Module):\n",
        "\n",
        "    def __init__(self, version='1_0', num_classes=1000):\n",
        "        super(SqueezeNet, self).__init__()\n",
        "        self.num_classes = num_classes\n",
        "        if version == '1_0':\n",
        "            self.features = nn.Sequential(\n",
        "                nn.Conv2d(3, 96, kernel_size=7, stride=2),\n",
        "                nn.ReLU(inplace=True),\n",
        "                nn.MaxPool2d(kernel_size=3, stride=2, ceil_mode=True),\n",
        "                Fire(96, 16, 64, 64),\n",
        "                Fire(128, 16, 64, 64),\n",
        "                Fire(128, 32, 128, 128),\n",
        "                nn.MaxPool2d(kernel_size=3, stride=2, ceil_mode=True),\n",
        "                Fire(256, 32, 128, 128),\n",
        "                Fire(256, 48, 192, 192),\n",
        "                Fire(384, 48, 192, 192),\n",
        "                Fire(384, 64, 256, 256),\n",
        "                nn.MaxPool2d(kernel_size=3, stride=2, ceil_mode=True),\n",
        "                Fire(512, 64, 256, 256),\n",
        "            )\n",
        "        elif version == '1_1':\n",
        "            self.features = nn.Sequential(\n",
        "                nn.Conv2d(3, 64, kernel_size=3, stride=2),\n",
        "                nn.ReLU(inplace=True),\n",
        "                nn.MaxPool2d(kernel_size=3, stride=2, ceil_mode=True),\n",
        "                Fire(64, 16, 64, 64),\n",
        "                Fire(128, 16, 64, 64),\n",
        "                nn.MaxPool2d(kernel_size=3, stride=2, ceil_mode=True),\n",
        "                Fire(128, 32, 128, 128),\n",
        "                Fire(256, 32, 128, 128),\n",
        "                nn.MaxPool2d(kernel_size=3, stride=2, ceil_mode=True),\n",
        "                Fire(256, 48, 192, 192),\n",
        "                Fire(384, 48, 192, 192),\n",
        "                Fire(384, 64, 256, 256),\n",
        "                Fire(512, 64, 256, 256),\n",
        "            )\n",
        "        else:\n",
        "            # FIXME: Is this needed? SqueezeNet should only be called from the\n",
        "            # FIXME: squeezenet1_x() functions\n",
        "            # FIXME: This checking is not done for the other models\n",
        "            raise ValueError(\"Unsupported SqueezeNet version {version}:\"\n",
        "                             \"1_0 or 1_1 expected\".format(version=version))\n",
        "\n",
        "        # Final convolution is initialized differently from the rest\n",
        "        final_conv = nn.Conv2d(512, self.num_classes, kernel_size=1)\n",
        "        self.classifier = nn.Sequential(\n",
        "            nn.Dropout(p=0.5),\n",
        "            final_conv,\n",
        "            nn.ReLU(inplace=True),\n",
        "            nn.AdaptiveAvgPool2d((1, 1))\n",
        "        )\n",
        "\n",
        "        for m in self.modules():\n",
        "            if isinstance(m, nn.Conv2d):\n",
        "                if m is final_conv:\n",
        "                    init.normal_(m.weight, mean=0.0, std=0.01)\n",
        "                else:\n",
        "                    init.kaiming_uniform_(m.weight)\n",
        "                if m.bias is not None:\n",
        "                    init.constant_(m.bias, 0)\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = self.features(x)\n",
        "        x = self.classifier(x)\n",
        "        return torch.flatten(x, 1)\n",
        "\n",
        "\n",
        "def _squeezenet(version, pretrained, progress, **kwargs):\n",
        "    model = SqueezeNet(version, **kwargs)\n",
        "    if pretrained:\n",
        "        arch = 'squeezenet' + version\n",
        "        state_dict = load_state_dict_from_url(model_urls[arch],\n",
        "                                              progress=progress)\n",
        "        model.load_state_dict(state_dict)\n",
        "    return model\n",
        "\n",
        "\n",
        "def squeezenet1_0(pretrained=False, progress=True, **kwargs):\n",
        "    r\"\"\"SqueezeNet model architecture from the `\"SqueezeNet: AlexNet-level\n",
        "    accuracy with 50x fewer parameters and <0.5MB model size\"\n",
        "    <https://arxiv.org/abs/1602.07360>`_ paper.\n",
        "    Args:\n",
        "        pretrained (bool): If True, returns a model pre-trained on ImageNet\n",
        "        progress (bool): If True, displays a progress bar of the download to stderr\n",
        "    \"\"\"\n",
        "    return _squeezenet('1_0', pretrained, progress, **kwargs)\n",
        "\n",
        "\n",
        "def squeezenet1_1(pretrained=False, progress=True, **kwargs):\n",
        "    r\"\"\"SqueezeNet 1.1 model from the `official SqueezeNet repo\n",
        "    <https://github.com/DeepScale/SqueezeNet/tree/master/SqueezeNet_v1.1>`_.\n",
        "    SqueezeNet 1.1 has 2.4x less computation and slightly fewer parameters\n",
        "    than SqueezeNet 1.0, without sacrificing accuracy.\n",
        "    Args:\n",
        "        pretrained (bool): If True, returns a model pre-trained on ImageNet\n",
        "        progress (bool): If True, displays a progress bar of the download to stderr\n",
        "    \"\"\"\n",
        "    return _squeezenet('1_1', pretrained, progress, **kwargs)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Xaz9jOeSZOVK",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "\n",
        "#change the line below for different models\n",
        "\n",
        "# import torchvision.models as models\n",
        "# model = models.squeezenet1_0(pretrained=True)\n",
        "\n",
        "model = squeezenet1_1(num_classes=10)\n",
        "\n",
        "# print(model)\n",
        "\n",
        "model = model.to(device=device, dtype=torch.float)\n",
        "\n",
        "# Cross Entropy Loss \n",
        "error = CrossEntropyLoss().to(device=device, dtype=torch.float)\n",
        "\n",
        "#Optimizer\n",
        "learning_rate = 0.1\n",
        "optimizer = adabound.AdaBound(model.parameters(), lr=1e-3, final_lr=learning_rate)\n",
        "\n",
        "#Optimizer adam\n",
        "# learning_rate = 0.04\n",
        "#optimizer = Adam(model.parameters(), lr=learning_rate)\n",
        "# optimizer = SGD(model.parameters(), lr=learning_rate, momentum=0.9, dampening=0, weight_decay=0.0002, nesterov=False)\n",
        "# optimizer = SGD(model.parameters(), lr=learning_rate)\n",
        "\n",
        "#adaptive learning rate\n",
        "# how many epoc before decreasing learning rate\n",
        "#exp_lr_scheduler = lr_scheduler.StepLR(optimizer, step_size=5, gamma=0.1)\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sCt69lxiZXZI",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#load trained dictionary\n",
        "#PATH = '/content/squeezenet_loss.pth'\n",
        "#model.load_state_dict(torch.load(PATH))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dT6xPT7cZZPY",
        "colab_type": "code",
        "outputId": "4425369d-dd46-4623-de9e-14a3d2500286",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "%%time\n",
        "#training/testing\n",
        "loss_list = []\n",
        "loss_list_test = []\n",
        "iteration_list = []\n",
        "accuracy_list = []\n",
        "# execution time checking\n",
        "execution_time = []\n",
        "\n",
        "PATH = '/content/squeezenet_acc.pth'\n",
        "best_model_wts = copy.deepcopy(model.state_dict())\n",
        "PATH2 = '/content/squeezenet_loss.pth'\n",
        "best_model_loss = copy.deepcopy(model.state_dict())\n",
        "\n",
        "#PATH = '/content/modified_mnist_effb2_pre_original_latest.pth'\n",
        "#latest_model_wts = copy.deepcopy(model.state_dict())\n",
        "best_loss = float(\"inf\")\n",
        "best_acc = 0.0\n",
        "#set number of epochs \n",
        "num_epochs = 100\n",
        "for epoch in range(num_epochs):\n",
        "    #print(\"\\nTrain Epoch {}: lr = {}\".format(epoch, exp_lr_scheduler.get_lr()[0]))\n",
        "    print(\"\\nTrain Epoch {}: lr = {}\".format(epoch,learning_rate))\n",
        "\n",
        "    start = timeit.default_timer()\n",
        "    loss_list.append(train(epoch, train_loader, model, error, optimizer, batch_size))\n",
        "    loss_test, accuracy = test(test_loader, model, error, batch_size)\n",
        "    stop = timeit.default_timer()\n",
        "    etime = stop - start\n",
        "    execution_time.append(etime)\n",
        "\n",
        "    loss_list_test.append(loss_test)\n",
        "    accuracy_list.append(accuracy)\n",
        "    iteration_list.append(epoch)\n",
        "    #exp_lr_scheduler.step()\n",
        "\n",
        "    #saving model with best acc \n",
        "    if accuracy > best_acc:\n",
        "      print(f\"Better accuracy at Epoch {epoch}: accuracy = {accuracy}%\")\n",
        "      best_acc = accuracy\n",
        "      best_model_wts = copy.deepcopy(model.state_dict())\n",
        "      torch.save(best_model_wts, PATH)\n",
        "    #saving model with best loss\n",
        "    if loss_test < best_loss:\n",
        "      print(f\"Better loss at Epoch {epoch}: loss = {loss_test}%\")\n",
        "      best_loss = loss_test\n",
        "      best_model_loss = copy.deepcopy(model.state_dict())\n",
        "      torch.save(best_model_loss, PATH2)\n",
        "#load model\n",
        "# model = Net()\n",
        "# model.load_state_dict(torch.load(PATH))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\n",
            "Train Epoch 0: lr = 0.1\n",
            "Train Epoch: 0 [5000/50000 (10%)]\tTrain Loss: 2.324690\n",
            "Train Epoch: 0 [10000/50000 (20%)]\tTrain Loss: 2.280486\n",
            "Train Epoch: 0 [15000/50000 (30%)]\tTrain Loss: 2.248495\n",
            "Train Epoch: 0 [20000/50000 (40%)]\tTrain Loss: 2.187538\n",
            "Train Epoch: 0 [25000/50000 (50%)]\tTrain Loss: 2.157347\n",
            "Train Epoch: 0 [30000/50000 (60%)]\tTrain Loss: 2.101729\n",
            "Train Epoch: 0 [35000/50000 (70%)]\tTrain Loss: 2.104143\n",
            "Train Epoch: 0 [40000/50000 (80%)]\tTrain Loss: 2.041676\n",
            "Train Epoch: 0 [45000/50000 (90%)]\tTrain Loss: 2.028886\n",
            "\n",
            "Test set: Test loss: 2.0140, Accuracy: 1365/5000 (27%)\n",
            "\n",
            "Better accuracy at Epoch 0: accuracy = 27.3%\n",
            "Better loss at Epoch 0: loss = 2.0140409195423126%\n",
            "\n",
            "Train Epoch 1: lr = 0.1\n",
            "Train Epoch: 1 [5000/50000 (10%)]\tTrain Loss: 2.009719\n",
            "Train Epoch: 1 [10000/50000 (20%)]\tTrain Loss: 1.970154\n",
            "Train Epoch: 1 [15000/50000 (30%)]\tTrain Loss: 1.942461\n",
            "Train Epoch: 1 [20000/50000 (40%)]\tTrain Loss: 1.952287\n",
            "Train Epoch: 1 [25000/50000 (50%)]\tTrain Loss: 1.940961\n",
            "Train Epoch: 1 [30000/50000 (60%)]\tTrain Loss: 1.927777\n",
            "Train Epoch: 1 [35000/50000 (70%)]\tTrain Loss: 1.932047\n",
            "Train Epoch: 1 [40000/50000 (80%)]\tTrain Loss: 1.921307\n",
            "Train Epoch: 1 [45000/50000 (90%)]\tTrain Loss: 1.902311\n",
            "\n",
            "Test set: Test loss: 1.8302, Accuracy: 1785/5000 (36%)\n",
            "\n",
            "Better accuracy at Epoch 1: accuracy = 35.7%\n",
            "Better loss at Epoch 1: loss = 1.830166883468628%\n",
            "\n",
            "Train Epoch 2: lr = 0.1\n",
            "Train Epoch: 2 [5000/50000 (10%)]\tTrain Loss: 1.893568\n",
            "Train Epoch: 2 [10000/50000 (20%)]\tTrain Loss: 1.839751\n",
            "Train Epoch: 2 [15000/50000 (30%)]\tTrain Loss: 1.831479\n",
            "Train Epoch: 2 [20000/50000 (40%)]\tTrain Loss: 1.834439\n",
            "Train Epoch: 2 [25000/50000 (50%)]\tTrain Loss: 1.858835\n",
            "Train Epoch: 2 [30000/50000 (60%)]\tTrain Loss: 1.827645\n",
            "Train Epoch: 2 [35000/50000 (70%)]\tTrain Loss: 1.847901\n",
            "Train Epoch: 2 [40000/50000 (80%)]\tTrain Loss: 1.809918\n",
            "Train Epoch: 2 [45000/50000 (90%)]\tTrain Loss: 1.792549\n",
            "\n",
            "Test set: Test loss: 1.8060, Accuracy: 1846/5000 (37%)\n",
            "\n",
            "Better accuracy at Epoch 2: accuracy = 36.92%\n",
            "Better loss at Epoch 2: loss = 1.8060061609745026%\n",
            "\n",
            "Train Epoch 3: lr = 0.1\n",
            "Train Epoch: 3 [5000/50000 (10%)]\tTrain Loss: 1.802737\n",
            "Train Epoch: 3 [10000/50000 (20%)]\tTrain Loss: 1.800918\n",
            "Train Epoch: 3 [15000/50000 (30%)]\tTrain Loss: 1.802704\n",
            "Train Epoch: 3 [20000/50000 (40%)]\tTrain Loss: 1.759598\n",
            "Train Epoch: 3 [25000/50000 (50%)]\tTrain Loss: 1.746764\n",
            "Train Epoch: 3 [30000/50000 (60%)]\tTrain Loss: 1.746486\n",
            "Train Epoch: 3 [35000/50000 (70%)]\tTrain Loss: 1.750349\n",
            "Train Epoch: 3 [40000/50000 (80%)]\tTrain Loss: 1.698980\n",
            "Train Epoch: 3 [45000/50000 (90%)]\tTrain Loss: 1.745168\n",
            "\n",
            "Test set: Test loss: 1.7108, Accuracy: 2185/5000 (44%)\n",
            "\n",
            "Better accuracy at Epoch 3: accuracy = 43.7%\n",
            "Better loss at Epoch 3: loss = 1.710838406085968%\n",
            "\n",
            "Train Epoch 4: lr = 0.1\n",
            "Train Epoch: 4 [5000/50000 (10%)]\tTrain Loss: 1.721554\n",
            "Train Epoch: 4 [10000/50000 (20%)]\tTrain Loss: 1.689682\n",
            "Train Epoch: 4 [15000/50000 (30%)]\tTrain Loss: 1.703244\n",
            "Train Epoch: 4 [20000/50000 (40%)]\tTrain Loss: 1.709314\n",
            "Train Epoch: 4 [25000/50000 (50%)]\tTrain Loss: 1.754006\n",
            "Train Epoch: 4 [30000/50000 (60%)]\tTrain Loss: 1.698532\n",
            "Train Epoch: 4 [35000/50000 (70%)]\tTrain Loss: 1.704872\n",
            "Train Epoch: 4 [40000/50000 (80%)]\tTrain Loss: 1.681801\n",
            "Train Epoch: 4 [45000/50000 (90%)]\tTrain Loss: 1.689907\n",
            "\n",
            "Test set: Test loss: 1.6433, Accuracy: 2184/5000 (44%)\n",
            "\n",
            "Better loss at Epoch 4: loss = 1.6432773303985595%\n",
            "\n",
            "Train Epoch 5: lr = 0.1\n",
            "Train Epoch: 5 [5000/50000 (10%)]\tTrain Loss: 1.685144\n",
            "Train Epoch: 5 [10000/50000 (20%)]\tTrain Loss: 1.683284\n",
            "Train Epoch: 5 [15000/50000 (30%)]\tTrain Loss: 1.690304\n",
            "Train Epoch: 5 [20000/50000 (40%)]\tTrain Loss: 1.699781\n",
            "Train Epoch: 5 [25000/50000 (50%)]\tTrain Loss: 1.655316\n",
            "Train Epoch: 5 [30000/50000 (60%)]\tTrain Loss: 1.660877\n",
            "Train Epoch: 5 [35000/50000 (70%)]\tTrain Loss: 1.665024\n",
            "Train Epoch: 5 [40000/50000 (80%)]\tTrain Loss: 1.630733\n",
            "Train Epoch: 5 [45000/50000 (90%)]\tTrain Loss: 1.676078\n",
            "\n",
            "Test set: Test loss: 1.6125, Accuracy: 2187/5000 (44%)\n",
            "\n",
            "Better accuracy at Epoch 5: accuracy = 43.74%\n",
            "Better loss at Epoch 5: loss = 1.612540248632431%\n",
            "\n",
            "Train Epoch 6: lr = 0.1\n",
            "Train Epoch: 6 [5000/50000 (10%)]\tTrain Loss: 1.622614\n",
            "Train Epoch: 6 [10000/50000 (20%)]\tTrain Loss: 1.549089\n",
            "Train Epoch: 6 [15000/50000 (30%)]\tTrain Loss: 1.559126\n",
            "Train Epoch: 6 [20000/50000 (40%)]\tTrain Loss: 1.555975\n",
            "Train Epoch: 6 [25000/50000 (50%)]\tTrain Loss: 1.518601\n",
            "Train Epoch: 6 [30000/50000 (60%)]\tTrain Loss: 1.524297\n",
            "Train Epoch: 6 [35000/50000 (70%)]\tTrain Loss: 1.522706\n",
            "Train Epoch: 6 [40000/50000 (80%)]\tTrain Loss: 1.521166\n",
            "Train Epoch: 6 [45000/50000 (90%)]\tTrain Loss: 1.516465\n",
            "\n",
            "Test set: Test loss: 1.4779, Accuracy: 2500/5000 (50%)\n",
            "\n",
            "Better accuracy at Epoch 6: accuracy = 50.0%\n",
            "Better loss at Epoch 6: loss = 1.4778608882427215%\n",
            "\n",
            "Train Epoch 7: lr = 0.1\n",
            "Train Epoch: 7 [5000/50000 (10%)]\tTrain Loss: 1.533004\n",
            "Train Epoch: 7 [10000/50000 (20%)]\tTrain Loss: 1.510467\n",
            "Train Epoch: 7 [15000/50000 (30%)]\tTrain Loss: 1.500793\n",
            "Train Epoch: 7 [20000/50000 (40%)]\tTrain Loss: 1.488992\n",
            "Train Epoch: 7 [25000/50000 (50%)]\tTrain Loss: 1.437714\n",
            "Train Epoch: 7 [30000/50000 (60%)]\tTrain Loss: 1.477619\n",
            "Train Epoch: 7 [35000/50000 (70%)]\tTrain Loss: 1.478457\n",
            "Train Epoch: 7 [40000/50000 (80%)]\tTrain Loss: 1.461223\n",
            "Train Epoch: 7 [45000/50000 (90%)]\tTrain Loss: 1.446909\n",
            "\n",
            "Test set: Test loss: 1.4461, Accuracy: 2591/5000 (52%)\n",
            "\n",
            "Better accuracy at Epoch 7: accuracy = 51.82%\n",
            "Better loss at Epoch 7: loss = 1.446122784614563%\n",
            "\n",
            "Train Epoch 8: lr = 0.1\n",
            "Train Epoch: 8 [5000/50000 (10%)]\tTrain Loss: 1.468824\n",
            "Train Epoch: 8 [10000/50000 (20%)]\tTrain Loss: 1.430019\n",
            "Train Epoch: 8 [15000/50000 (30%)]\tTrain Loss: 1.427867\n",
            "Train Epoch: 8 [20000/50000 (40%)]\tTrain Loss: 1.456425\n",
            "Train Epoch: 8 [25000/50000 (50%)]\tTrain Loss: 1.458756\n",
            "Train Epoch: 8 [30000/50000 (60%)]\tTrain Loss: 1.411486\n",
            "Train Epoch: 8 [35000/50000 (70%)]\tTrain Loss: 1.458871\n",
            "Train Epoch: 8 [40000/50000 (80%)]\tTrain Loss: 1.445652\n",
            "Train Epoch: 8 [45000/50000 (90%)]\tTrain Loss: 1.423023\n",
            "\n",
            "Test set: Test loss: 1.4552, Accuracy: 2501/5000 (50%)\n",
            "\n",
            "\n",
            "Train Epoch 9: lr = 0.1\n",
            "Train Epoch: 9 [5000/50000 (10%)]\tTrain Loss: 1.450594\n",
            "Train Epoch: 9 [10000/50000 (20%)]\tTrain Loss: 1.386581\n",
            "Train Epoch: 9 [15000/50000 (30%)]\tTrain Loss: 1.421920\n",
            "Train Epoch: 9 [20000/50000 (40%)]\tTrain Loss: 1.431581\n",
            "Train Epoch: 9 [25000/50000 (50%)]\tTrain Loss: 1.408708\n",
            "Train Epoch: 9 [30000/50000 (60%)]\tTrain Loss: 1.408633\n",
            "Train Epoch: 9 [35000/50000 (70%)]\tTrain Loss: 1.424633\n",
            "Train Epoch: 9 [40000/50000 (80%)]\tTrain Loss: 1.395047\n",
            "Train Epoch: 9 [45000/50000 (90%)]\tTrain Loss: 1.403522\n",
            "\n",
            "Test set: Test loss: 1.4099, Accuracy: 2549/5000 (51%)\n",
            "\n",
            "Better loss at Epoch 9: loss = 1.409937480688095%\n",
            "\n",
            "Train Epoch 10: lr = 0.1\n",
            "Train Epoch: 10 [5000/50000 (10%)]\tTrain Loss: 1.388796\n",
            "Train Epoch: 10 [10000/50000 (20%)]\tTrain Loss: 1.395948\n",
            "Train Epoch: 10 [15000/50000 (30%)]\tTrain Loss: 1.353834\n",
            "Train Epoch: 10 [20000/50000 (40%)]\tTrain Loss: 1.391955\n",
            "Train Epoch: 10 [25000/50000 (50%)]\tTrain Loss: 1.387935\n",
            "Train Epoch: 10 [30000/50000 (60%)]\tTrain Loss: 1.381369\n",
            "Train Epoch: 10 [35000/50000 (70%)]\tTrain Loss: 1.409187\n",
            "Train Epoch: 10 [40000/50000 (80%)]\tTrain Loss: 1.404000\n",
            "Train Epoch: 10 [45000/50000 (90%)]\tTrain Loss: 1.397999\n",
            "\n",
            "Test set: Test loss: 1.4159, Accuracy: 2613/5000 (52%)\n",
            "\n",
            "Better accuracy at Epoch 10: accuracy = 52.26%\n",
            "\n",
            "Train Epoch 11: lr = 0.1\n",
            "Train Epoch: 11 [5000/50000 (10%)]\tTrain Loss: 1.368649\n",
            "Train Epoch: 11 [10000/50000 (20%)]\tTrain Loss: 1.343765\n",
            "Train Epoch: 11 [15000/50000 (30%)]\tTrain Loss: 1.366442\n",
            "Train Epoch: 11 [20000/50000 (40%)]\tTrain Loss: 1.362393\n",
            "Train Epoch: 11 [25000/50000 (50%)]\tTrain Loss: 1.388599\n",
            "Train Epoch: 11 [30000/50000 (60%)]\tTrain Loss: 1.346965\n",
            "Train Epoch: 11 [35000/50000 (70%)]\tTrain Loss: 1.376698\n",
            "Train Epoch: 11 [40000/50000 (80%)]\tTrain Loss: 1.380929\n",
            "Train Epoch: 11 [45000/50000 (90%)]\tTrain Loss: 1.328202\n",
            "\n",
            "Test set: Test loss: 1.3654, Accuracy: 2692/5000 (54%)\n",
            "\n",
            "Better accuracy at Epoch 11: accuracy = 53.84%\n",
            "Better loss at Epoch 11: loss = 1.3654066479206086%\n",
            "\n",
            "Train Epoch 12: lr = 0.1\n",
            "Train Epoch: 12 [5000/50000 (10%)]\tTrain Loss: 1.353262\n",
            "Train Epoch: 12 [10000/50000 (20%)]\tTrain Loss: 1.371139\n",
            "Train Epoch: 12 [15000/50000 (30%)]\tTrain Loss: 1.334590\n",
            "Train Epoch: 12 [20000/50000 (40%)]\tTrain Loss: 1.373329\n",
            "Train Epoch: 12 [25000/50000 (50%)]\tTrain Loss: 1.346873\n",
            "Train Epoch: 12 [30000/50000 (60%)]\tTrain Loss: 1.346647\n",
            "Train Epoch: 12 [35000/50000 (70%)]\tTrain Loss: 1.339755\n",
            "Train Epoch: 12 [40000/50000 (80%)]\tTrain Loss: 1.351279\n",
            "Train Epoch: 12 [45000/50000 (90%)]\tTrain Loss: 1.344150\n",
            "\n",
            "Test set: Test loss: 1.3537, Accuracy: 2724/5000 (54%)\n",
            "\n",
            "Better accuracy at Epoch 12: accuracy = 54.48%\n",
            "Better loss at Epoch 12: loss = 1.3537101686000823%\n",
            "\n",
            "Train Epoch 13: lr = 0.1\n",
            "Train Epoch: 13 [5000/50000 (10%)]\tTrain Loss: 1.318021\n",
            "Train Epoch: 13 [10000/50000 (20%)]\tTrain Loss: 1.353465\n",
            "Train Epoch: 13 [15000/50000 (30%)]\tTrain Loss: 1.324393\n",
            "Train Epoch: 13 [20000/50000 (40%)]\tTrain Loss: 1.299415\n",
            "Train Epoch: 13 [25000/50000 (50%)]\tTrain Loss: 1.326412\n",
            "Train Epoch: 13 [30000/50000 (60%)]\tTrain Loss: 1.308008\n",
            "Train Epoch: 13 [35000/50000 (70%)]\tTrain Loss: 1.366003\n",
            "Train Epoch: 13 [40000/50000 (80%)]\tTrain Loss: 1.305651\n",
            "Train Epoch: 13 [45000/50000 (90%)]\tTrain Loss: 1.337771\n",
            "\n",
            "Test set: Test loss: 1.4223, Accuracy: 2565/5000 (51%)\n",
            "\n",
            "\n",
            "Train Epoch 14: lr = 0.1\n",
            "Train Epoch: 14 [5000/50000 (10%)]\tTrain Loss: 1.293460\n",
            "Train Epoch: 14 [10000/50000 (20%)]\tTrain Loss: 1.304548\n",
            "Train Epoch: 14 [15000/50000 (30%)]\tTrain Loss: 1.293581\n",
            "Train Epoch: 14 [20000/50000 (40%)]\tTrain Loss: 1.334519\n",
            "Train Epoch: 14 [25000/50000 (50%)]\tTrain Loss: 1.308264\n",
            "Train Epoch: 14 [30000/50000 (60%)]\tTrain Loss: 1.315960\n",
            "Train Epoch: 14 [35000/50000 (70%)]\tTrain Loss: 1.304214\n",
            "Train Epoch: 14 [40000/50000 (80%)]\tTrain Loss: 1.306800\n",
            "Train Epoch: 14 [45000/50000 (90%)]\tTrain Loss: 1.309767\n",
            "\n",
            "Test set: Test loss: 1.3546, Accuracy: 2754/5000 (55%)\n",
            "\n",
            "Better accuracy at Epoch 14: accuracy = 55.08%\n",
            "\n",
            "Train Epoch 15: lr = 0.1\n",
            "Train Epoch: 15 [5000/50000 (10%)]\tTrain Loss: 1.337493\n",
            "Train Epoch: 15 [10000/50000 (20%)]\tTrain Loss: 1.294535\n",
            "Train Epoch: 15 [15000/50000 (30%)]\tTrain Loss: 1.278667\n",
            "Train Epoch: 15 [20000/50000 (40%)]\tTrain Loss: 1.312302\n",
            "Train Epoch: 15 [25000/50000 (50%)]\tTrain Loss: 1.275536\n",
            "Train Epoch: 15 [30000/50000 (60%)]\tTrain Loss: 1.260803\n",
            "Train Epoch: 15 [35000/50000 (70%)]\tTrain Loss: 1.278191\n",
            "Train Epoch: 15 [40000/50000 (80%)]\tTrain Loss: 1.238726\n",
            "Train Epoch: 15 [45000/50000 (90%)]\tTrain Loss: 1.283534\n",
            "\n",
            "Test set: Test loss: 1.3036, Accuracy: 2899/5000 (58%)\n",
            "\n",
            "Better accuracy at Epoch 15: accuracy = 57.98%\n",
            "Better loss at Epoch 15: loss = 1.3036024570465088%\n",
            "\n",
            "Train Epoch 16: lr = 0.1\n",
            "Train Epoch: 16 [5000/50000 (10%)]\tTrain Loss: 1.251678\n",
            "Train Epoch: 16 [10000/50000 (20%)]\tTrain Loss: 1.226897\n",
            "Train Epoch: 16 [15000/50000 (30%)]\tTrain Loss: 1.240211\n",
            "Train Epoch: 16 [20000/50000 (40%)]\tTrain Loss: 1.240072\n",
            "Train Epoch: 16 [25000/50000 (50%)]\tTrain Loss: 1.262142\n",
            "Train Epoch: 16 [30000/50000 (60%)]\tTrain Loss: 1.250740\n",
            "Train Epoch: 16 [35000/50000 (70%)]\tTrain Loss: 1.210794\n",
            "Train Epoch: 16 [40000/50000 (80%)]\tTrain Loss: 1.270835\n",
            "Train Epoch: 16 [45000/50000 (90%)]\tTrain Loss: 1.236724\n",
            "\n",
            "Test set: Test loss: 1.2585, Accuracy: 2915/5000 (58%)\n",
            "\n",
            "Better accuracy at Epoch 16: accuracy = 58.3%\n",
            "Better loss at Epoch 16: loss = 1.2585174471139908%\n",
            "\n",
            "Train Epoch 17: lr = 0.1\n",
            "Train Epoch: 17 [5000/50000 (10%)]\tTrain Loss: 1.232621\n",
            "Train Epoch: 17 [10000/50000 (20%)]\tTrain Loss: 1.206081\n",
            "Train Epoch: 17 [15000/50000 (30%)]\tTrain Loss: 1.177945\n",
            "Train Epoch: 17 [20000/50000 (40%)]\tTrain Loss: 1.210133\n",
            "Train Epoch: 17 [25000/50000 (50%)]\tTrain Loss: 1.207818\n",
            "Train Epoch: 17 [30000/50000 (60%)]\tTrain Loss: 1.207524\n",
            "Train Epoch: 17 [35000/50000 (70%)]\tTrain Loss: 1.231288\n",
            "Train Epoch: 17 [40000/50000 (80%)]\tTrain Loss: 1.234526\n",
            "Train Epoch: 17 [45000/50000 (90%)]\tTrain Loss: 1.239544\n",
            "\n",
            "Test set: Test loss: 1.3263, Accuracy: 2899/5000 (58%)\n",
            "\n",
            "\n",
            "Train Epoch 18: lr = 0.1\n",
            "Train Epoch: 18 [5000/50000 (10%)]\tTrain Loss: 1.218373\n",
            "Train Epoch: 18 [10000/50000 (20%)]\tTrain Loss: 1.190467\n",
            "Train Epoch: 18 [15000/50000 (30%)]\tTrain Loss: 1.201526\n",
            "Train Epoch: 18 [20000/50000 (40%)]\tTrain Loss: 1.239757\n",
            "Train Epoch: 18 [25000/50000 (50%)]\tTrain Loss: 1.204345\n",
            "Train Epoch: 18 [30000/50000 (60%)]\tTrain Loss: 1.200679\n",
            "Train Epoch: 18 [35000/50000 (70%)]\tTrain Loss: 1.191683\n",
            "Train Epoch: 18 [40000/50000 (80%)]\tTrain Loss: 1.221592\n",
            "Train Epoch: 18 [45000/50000 (90%)]\tTrain Loss: 1.103492\n",
            "\n",
            "Test set: Test loss: 1.1963, Accuracy: 2963/5000 (59%)\n",
            "\n",
            "Better accuracy at Epoch 18: accuracy = 59.26%\n",
            "Better loss at Epoch 18: loss = 1.196286724805832%\n",
            "\n",
            "Train Epoch 19: lr = 0.1\n",
            "Train Epoch: 19 [5000/50000 (10%)]\tTrain Loss: 1.151181\n",
            "Train Epoch: 19 [10000/50000 (20%)]\tTrain Loss: 1.063829\n",
            "Train Epoch: 19 [15000/50000 (30%)]\tTrain Loss: 1.117427\n",
            "Train Epoch: 19 [20000/50000 (40%)]\tTrain Loss: 1.173038\n",
            "Train Epoch: 19 [25000/50000 (50%)]\tTrain Loss: 1.086487\n",
            "Train Epoch: 19 [30000/50000 (60%)]\tTrain Loss: 1.091895\n",
            "Train Epoch: 19 [35000/50000 (70%)]\tTrain Loss: 1.062531\n",
            "Train Epoch: 19 [40000/50000 (80%)]\tTrain Loss: 1.093799\n",
            "Train Epoch: 19 [45000/50000 (90%)]\tTrain Loss: 1.117538\n",
            "\n",
            "Test set: Test loss: 1.1258, Accuracy: 3085/5000 (62%)\n",
            "\n",
            "Better accuracy at Epoch 19: accuracy = 61.7%\n",
            "Better loss at Epoch 19: loss = 1.125831394791603%\n",
            "\n",
            "Train Epoch 20: lr = 0.1\n",
            "Train Epoch: 20 [5000/50000 (10%)]\tTrain Loss: 1.089629\n",
            "Train Epoch: 20 [10000/50000 (20%)]\tTrain Loss: 1.028529\n",
            "Train Epoch: 20 [15000/50000 (30%)]\tTrain Loss: 1.065910\n",
            "Train Epoch: 20 [20000/50000 (40%)]\tTrain Loss: 1.055044\n",
            "Train Epoch: 20 [25000/50000 (50%)]\tTrain Loss: 1.042082\n",
            "Train Epoch: 20 [30000/50000 (60%)]\tTrain Loss: 1.072607\n",
            "Train Epoch: 20 [35000/50000 (70%)]\tTrain Loss: 1.052812\n",
            "Train Epoch: 20 [40000/50000 (80%)]\tTrain Loss: 1.044043\n",
            "Train Epoch: 20 [45000/50000 (90%)]\tTrain Loss: 1.048756\n",
            "\n",
            "Test set: Test loss: 1.1110, Accuracy: 3126/5000 (63%)\n",
            "\n",
            "Better accuracy at Epoch 20: accuracy = 62.52%\n",
            "Better loss at Epoch 20: loss = 1.111008328795433%\n",
            "\n",
            "Train Epoch 21: lr = 0.1\n",
            "Train Epoch: 21 [5000/50000 (10%)]\tTrain Loss: 1.045316\n",
            "Train Epoch: 21 [10000/50000 (20%)]\tTrain Loss: 1.032498\n",
            "Train Epoch: 21 [15000/50000 (30%)]\tTrain Loss: 1.053528\n",
            "Train Epoch: 21 [20000/50000 (40%)]\tTrain Loss: 1.028824\n",
            "Train Epoch: 21 [25000/50000 (50%)]\tTrain Loss: 1.044861\n",
            "Train Epoch: 21 [30000/50000 (60%)]\tTrain Loss: 1.028661\n",
            "Train Epoch: 21 [35000/50000 (70%)]\tTrain Loss: 1.038669\n",
            "Train Epoch: 21 [40000/50000 (80%)]\tTrain Loss: 1.076446\n",
            "Train Epoch: 21 [45000/50000 (90%)]\tTrain Loss: 1.076377\n",
            "\n",
            "Test set: Test loss: 1.1431, Accuracy: 3090/5000 (62%)\n",
            "\n",
            "\n",
            "Train Epoch 22: lr = 0.1\n",
            "Train Epoch: 22 [5000/50000 (10%)]\tTrain Loss: 1.063215\n",
            "Train Epoch: 22 [10000/50000 (20%)]\tTrain Loss: 1.063794\n",
            "Train Epoch: 22 [15000/50000 (30%)]\tTrain Loss: 1.062148\n",
            "Train Epoch: 22 [20000/50000 (40%)]\tTrain Loss: 1.020637\n",
            "Train Epoch: 22 [25000/50000 (50%)]\tTrain Loss: 1.062070\n",
            "Train Epoch: 22 [30000/50000 (60%)]\tTrain Loss: 1.014032\n",
            "Train Epoch: 22 [35000/50000 (70%)]\tTrain Loss: 1.015267\n",
            "Train Epoch: 22 [40000/50000 (80%)]\tTrain Loss: 1.045125\n",
            "Train Epoch: 22 [45000/50000 (90%)]\tTrain Loss: 1.012459\n",
            "\n",
            "Test set: Test loss: 1.0602, Accuracy: 3216/5000 (64%)\n",
            "\n",
            "Better accuracy at Epoch 22: accuracy = 64.32%\n",
            "Better loss at Epoch 22: loss = 1.060160037279129%\n",
            "\n",
            "Train Epoch 23: lr = 0.1\n",
            "Train Epoch: 23 [5000/50000 (10%)]\tTrain Loss: 0.998828\n",
            "Train Epoch: 23 [10000/50000 (20%)]\tTrain Loss: 1.015623\n",
            "Train Epoch: 23 [15000/50000 (30%)]\tTrain Loss: 0.987195\n",
            "Train Epoch: 23 [20000/50000 (40%)]\tTrain Loss: 0.984016\n",
            "Train Epoch: 23 [25000/50000 (50%)]\tTrain Loss: 0.998608\n",
            "Train Epoch: 23 [30000/50000 (60%)]\tTrain Loss: 1.025086\n",
            "Train Epoch: 23 [35000/50000 (70%)]\tTrain Loss: 0.993781\n",
            "Train Epoch: 23 [40000/50000 (80%)]\tTrain Loss: 0.998136\n",
            "Train Epoch: 23 [45000/50000 (90%)]\tTrain Loss: 1.072821\n",
            "\n",
            "Test set: Test loss: 1.1359, Accuracy: 3199/5000 (64%)\n",
            "\n",
            "\n",
            "Train Epoch 24: lr = 0.1\n",
            "Train Epoch: 24 [5000/50000 (10%)]\tTrain Loss: 0.981858\n",
            "Train Epoch: 24 [10000/50000 (20%)]\tTrain Loss: 0.995260\n",
            "Train Epoch: 24 [15000/50000 (30%)]\tTrain Loss: 1.010875\n",
            "Train Epoch: 24 [20000/50000 (40%)]\tTrain Loss: 0.988845\n",
            "Train Epoch: 24 [25000/50000 (50%)]\tTrain Loss: 1.036600\n",
            "Train Epoch: 24 [30000/50000 (60%)]\tTrain Loss: 0.969025\n",
            "Train Epoch: 24 [35000/50000 (70%)]\tTrain Loss: 0.996260\n",
            "Train Epoch: 24 [40000/50000 (80%)]\tTrain Loss: 0.989422\n",
            "Train Epoch: 24 [45000/50000 (90%)]\tTrain Loss: 1.064958\n",
            "\n",
            "Test set: Test loss: 1.0992, Accuracy: 3151/5000 (63%)\n",
            "\n",
            "\n",
            "Train Epoch 25: lr = 0.1\n",
            "Train Epoch: 25 [5000/50000 (10%)]\tTrain Loss: 1.005634\n",
            "Train Epoch: 25 [10000/50000 (20%)]\tTrain Loss: 0.979135\n",
            "Train Epoch: 25 [15000/50000 (30%)]\tTrain Loss: 1.012196\n",
            "Train Epoch: 25 [20000/50000 (40%)]\tTrain Loss: 0.962940\n",
            "Train Epoch: 25 [25000/50000 (50%)]\tTrain Loss: 0.976827\n",
            "Train Epoch: 25 [30000/50000 (60%)]\tTrain Loss: 0.996005\n",
            "Train Epoch: 25 [35000/50000 (70%)]\tTrain Loss: 1.000301\n",
            "Train Epoch: 25 [40000/50000 (80%)]\tTrain Loss: 1.002287\n",
            "Train Epoch: 25 [45000/50000 (90%)]\tTrain Loss: 1.039881\n",
            "\n",
            "Test set: Test loss: 1.0686, Accuracy: 3201/5000 (64%)\n",
            "\n",
            "\n",
            "Train Epoch 26: lr = 0.1\n",
            "Train Epoch: 26 [5000/50000 (10%)]\tTrain Loss: 0.982978\n",
            "Train Epoch: 26 [10000/50000 (20%)]\tTrain Loss: 0.979067\n",
            "Train Epoch: 26 [15000/50000 (30%)]\tTrain Loss: 0.971096\n",
            "Train Epoch: 26 [20000/50000 (40%)]\tTrain Loss: 0.966586\n",
            "Train Epoch: 26 [25000/50000 (50%)]\tTrain Loss: 0.965291\n",
            "Train Epoch: 26 [30000/50000 (60%)]\tTrain Loss: 1.003262\n",
            "Train Epoch: 26 [35000/50000 (70%)]\tTrain Loss: 0.981691\n",
            "Train Epoch: 26 [40000/50000 (80%)]\tTrain Loss: 0.957406\n",
            "Train Epoch: 26 [45000/50000 (90%)]\tTrain Loss: 0.988815\n",
            "\n",
            "Test set: Test loss: 1.1067, Accuracy: 3212/5000 (64%)\n",
            "\n",
            "\n",
            "Train Epoch 27: lr = 0.1\n",
            "Train Epoch: 27 [5000/50000 (10%)]\tTrain Loss: 0.953165\n",
            "Train Epoch: 27 [10000/50000 (20%)]\tTrain Loss: 0.960741\n",
            "Train Epoch: 27 [15000/50000 (30%)]\tTrain Loss: 0.979153\n",
            "Train Epoch: 27 [20000/50000 (40%)]\tTrain Loss: 0.929990\n",
            "Train Epoch: 27 [25000/50000 (50%)]\tTrain Loss: 1.017865\n",
            "Train Epoch: 27 [30000/50000 (60%)]\tTrain Loss: 0.940864\n",
            "Train Epoch: 27 [35000/50000 (70%)]\tTrain Loss: 1.027553\n",
            "Train Epoch: 27 [40000/50000 (80%)]\tTrain Loss: 0.991308\n",
            "Train Epoch: 27 [45000/50000 (90%)]\tTrain Loss: 0.963638\n",
            "\n",
            "Test set: Test loss: 1.0774, Accuracy: 3222/5000 (64%)\n",
            "\n",
            "Better accuracy at Epoch 27: accuracy = 64.44%\n",
            "\n",
            "Train Epoch 28: lr = 0.1\n",
            "Train Epoch: 28 [5000/50000 (10%)]\tTrain Loss: 0.983876\n",
            "Train Epoch: 28 [10000/50000 (20%)]\tTrain Loss: 0.971749\n",
            "Train Epoch: 28 [15000/50000 (30%)]\tTrain Loss: 0.980954\n",
            "Train Epoch: 28 [20000/50000 (40%)]\tTrain Loss: 0.961373\n",
            "Train Epoch: 28 [25000/50000 (50%)]\tTrain Loss: 0.943794\n",
            "Train Epoch: 28 [30000/50000 (60%)]\tTrain Loss: 0.982785\n",
            "Train Epoch: 28 [35000/50000 (70%)]\tTrain Loss: 0.973089\n",
            "Train Epoch: 28 [40000/50000 (80%)]\tTrain Loss: 0.953860\n",
            "Train Epoch: 28 [45000/50000 (90%)]\tTrain Loss: 1.043562\n",
            "\n",
            "Test set: Test loss: 1.1904, Accuracy: 3075/5000 (62%)\n",
            "\n",
            "\n",
            "Train Epoch 29: lr = 0.1\n",
            "Train Epoch: 29 [5000/50000 (10%)]\tTrain Loss: 0.939844\n",
            "Train Epoch: 29 [10000/50000 (20%)]\tTrain Loss: 0.920958\n",
            "Train Epoch: 29 [15000/50000 (30%)]\tTrain Loss: 0.910149\n",
            "Train Epoch: 29 [20000/50000 (40%)]\tTrain Loss: 0.990273\n",
            "Train Epoch: 29 [25000/50000 (50%)]\tTrain Loss: 0.978566\n",
            "Train Epoch: 29 [30000/50000 (60%)]\tTrain Loss: 0.976039\n",
            "Train Epoch: 29 [35000/50000 (70%)]\tTrain Loss: 0.956197\n",
            "Train Epoch: 29 [40000/50000 (80%)]\tTrain Loss: 0.971861\n",
            "Train Epoch: 29 [45000/50000 (90%)]\tTrain Loss: 0.971926\n",
            "\n",
            "Test set: Test loss: 1.0022, Accuracy: 3297/5000 (66%)\n",
            "\n",
            "Better accuracy at Epoch 29: accuracy = 65.94%\n",
            "Better loss at Epoch 29: loss = 1.0022376346588135%\n",
            "\n",
            "Train Epoch 30: lr = 0.1\n",
            "Train Epoch: 30 [5000/50000 (10%)]\tTrain Loss: 0.945108\n",
            "Train Epoch: 30 [10000/50000 (20%)]\tTrain Loss: 0.921039\n",
            "Train Epoch: 30 [15000/50000 (30%)]\tTrain Loss: 0.933380\n",
            "Train Epoch: 30 [20000/50000 (40%)]\tTrain Loss: 0.953639\n",
            "Train Epoch: 30 [25000/50000 (50%)]\tTrain Loss: 0.980671\n",
            "Train Epoch: 30 [30000/50000 (60%)]\tTrain Loss: 0.936754\n",
            "Train Epoch: 30 [35000/50000 (70%)]\tTrain Loss: 0.957846\n",
            "Train Epoch: 30 [40000/50000 (80%)]\tTrain Loss: 0.946972\n",
            "Train Epoch: 30 [45000/50000 (90%)]\tTrain Loss: 1.050895\n",
            "\n",
            "Test set: Test loss: 1.0246, Accuracy: 3294/5000 (66%)\n",
            "\n",
            "\n",
            "Train Epoch 31: lr = 0.1\n",
            "Train Epoch: 31 [5000/50000 (10%)]\tTrain Loss: 0.957051\n",
            "Train Epoch: 31 [10000/50000 (20%)]\tTrain Loss: 0.935065\n",
            "Train Epoch: 31 [15000/50000 (30%)]\tTrain Loss: 0.921280\n",
            "Train Epoch: 31 [20000/50000 (40%)]\tTrain Loss: 0.961047\n",
            "Train Epoch: 31 [25000/50000 (50%)]\tTrain Loss: 0.951929\n",
            "Train Epoch: 31 [30000/50000 (60%)]\tTrain Loss: 0.976374\n",
            "Train Epoch: 31 [35000/50000 (70%)]\tTrain Loss: 0.954802\n",
            "Train Epoch: 31 [40000/50000 (80%)]\tTrain Loss: 0.924494\n",
            "Train Epoch: 31 [45000/50000 (90%)]\tTrain Loss: 0.907606\n",
            "\n",
            "Test set: Test loss: 1.0635, Accuracy: 3313/5000 (66%)\n",
            "\n",
            "Better accuracy at Epoch 31: accuracy = 66.26%\n",
            "\n",
            "Train Epoch 32: lr = 0.1\n",
            "Train Epoch: 32 [5000/50000 (10%)]\tTrain Loss: 0.908598\n",
            "Train Epoch: 32 [10000/50000 (20%)]\tTrain Loss: 0.903291\n",
            "Train Epoch: 32 [15000/50000 (30%)]\tTrain Loss: 0.902283\n",
            "Train Epoch: 32 [20000/50000 (40%)]\tTrain Loss: 0.962302\n",
            "Train Epoch: 32 [25000/50000 (50%)]\tTrain Loss: 0.906786\n",
            "Train Epoch: 32 [30000/50000 (60%)]\tTrain Loss: 0.938015\n",
            "Train Epoch: 32 [35000/50000 (70%)]\tTrain Loss: 0.947367\n",
            "Train Epoch: 32 [40000/50000 (80%)]\tTrain Loss: 0.971109\n",
            "Train Epoch: 32 [45000/50000 (90%)]\tTrain Loss: 0.967127\n",
            "\n",
            "Test set: Test loss: 1.0310, Accuracy: 3298/5000 (66%)\n",
            "\n",
            "\n",
            "Train Epoch 33: lr = 0.1\n",
            "Train Epoch: 33 [5000/50000 (10%)]\tTrain Loss: 0.956524\n",
            "Train Epoch: 33 [10000/50000 (20%)]\tTrain Loss: 0.980341\n",
            "Train Epoch: 33 [15000/50000 (30%)]\tTrain Loss: 1.004671\n",
            "Train Epoch: 33 [20000/50000 (40%)]\tTrain Loss: 0.930862\n",
            "Train Epoch: 33 [25000/50000 (50%)]\tTrain Loss: 0.946567\n",
            "Train Epoch: 33 [30000/50000 (60%)]\tTrain Loss: 0.950894\n",
            "Train Epoch: 33 [35000/50000 (70%)]\tTrain Loss: 0.952470\n",
            "Train Epoch: 33 [40000/50000 (80%)]\tTrain Loss: 0.963946\n",
            "Train Epoch: 33 [45000/50000 (90%)]\tTrain Loss: 0.990497\n",
            "\n",
            "Test set: Test loss: 1.0685, Accuracy: 3255/5000 (65%)\n",
            "\n",
            "\n",
            "Train Epoch 34: lr = 0.1\n",
            "Train Epoch: 34 [5000/50000 (10%)]\tTrain Loss: 0.944096\n",
            "Train Epoch: 34 [10000/50000 (20%)]\tTrain Loss: 0.920614\n",
            "Train Epoch: 34 [15000/50000 (30%)]\tTrain Loss: 0.967557\n",
            "Train Epoch: 34 [20000/50000 (40%)]\tTrain Loss: 0.948623\n",
            "Train Epoch: 34 [25000/50000 (50%)]\tTrain Loss: 1.011179\n",
            "Train Epoch: 34 [30000/50000 (60%)]\tTrain Loss: 0.895538\n",
            "Train Epoch: 34 [35000/50000 (70%)]\tTrain Loss: 1.024259\n",
            "Train Epoch: 34 [40000/50000 (80%)]\tTrain Loss: 0.971171\n",
            "Train Epoch: 34 [45000/50000 (90%)]\tTrain Loss: 0.941254\n",
            "\n",
            "Test set: Test loss: 0.9941, Accuracy: 3372/5000 (67%)\n",
            "\n",
            "Better accuracy at Epoch 34: accuracy = 67.44%\n",
            "Better loss at Epoch 34: loss = 0.9940930634737015%\n",
            "\n",
            "Train Epoch 35: lr = 0.1\n",
            "Train Epoch: 35 [5000/50000 (10%)]\tTrain Loss: 0.905283\n",
            "Train Epoch: 35 [10000/50000 (20%)]\tTrain Loss: 0.906681\n",
            "Train Epoch: 35 [15000/50000 (30%)]\tTrain Loss: 0.935079\n",
            "Train Epoch: 35 [20000/50000 (40%)]\tTrain Loss: 0.929535\n",
            "Train Epoch: 35 [25000/50000 (50%)]\tTrain Loss: 0.919881\n",
            "Train Epoch: 35 [30000/50000 (60%)]\tTrain Loss: 0.928806\n",
            "Train Epoch: 35 [35000/50000 (70%)]\tTrain Loss: 0.941351\n",
            "Train Epoch: 35 [40000/50000 (80%)]\tTrain Loss: 0.880761\n",
            "Train Epoch: 35 [45000/50000 (90%)]\tTrain Loss: 0.980235\n",
            "\n",
            "Test set: Test loss: 0.9593, Accuracy: 3440/5000 (69%)\n",
            "\n",
            "Better accuracy at Epoch 35: accuracy = 68.8%\n",
            "Better loss at Epoch 35: loss = 0.9593399733304977%\n",
            "\n",
            "Train Epoch 36: lr = 0.1\n",
            "Train Epoch: 36 [5000/50000 (10%)]\tTrain Loss: 0.860119\n",
            "Train Epoch: 36 [10000/50000 (20%)]\tTrain Loss: 0.864517\n",
            "Train Epoch: 36 [15000/50000 (30%)]\tTrain Loss: 0.914182\n",
            "Train Epoch: 36 [20000/50000 (40%)]\tTrain Loss: 0.935418\n",
            "Train Epoch: 36 [25000/50000 (50%)]\tTrain Loss: 0.919993\n",
            "Train Epoch: 36 [30000/50000 (60%)]\tTrain Loss: 0.934538\n",
            "Train Epoch: 36 [35000/50000 (70%)]\tTrain Loss: 0.898705\n",
            "Train Epoch: 36 [40000/50000 (80%)]\tTrain Loss: 0.938172\n",
            "Train Epoch: 36 [45000/50000 (90%)]\tTrain Loss: 0.985897\n",
            "\n",
            "Test set: Test loss: 0.9600, Accuracy: 3422/5000 (68%)\n",
            "\n",
            "\n",
            "Train Epoch 37: lr = 0.1\n",
            "Train Epoch: 37 [5000/50000 (10%)]\tTrain Loss: 0.893053\n",
            "Train Epoch: 37 [10000/50000 (20%)]\tTrain Loss: 0.889352\n",
            "Train Epoch: 37 [15000/50000 (30%)]\tTrain Loss: 0.898288\n",
            "Train Epoch: 37 [20000/50000 (40%)]\tTrain Loss: 0.916036\n",
            "Train Epoch: 37 [25000/50000 (50%)]\tTrain Loss: 0.902340\n",
            "Train Epoch: 37 [30000/50000 (60%)]\tTrain Loss: 0.901251\n",
            "Train Epoch: 37 [35000/50000 (70%)]\tTrain Loss: 0.931805\n",
            "Train Epoch: 37 [40000/50000 (80%)]\tTrain Loss: 0.898684\n",
            "Train Epoch: 37 [45000/50000 (90%)]\tTrain Loss: 0.949205\n",
            "\n",
            "Test set: Test loss: 1.0445, Accuracy: 3260/5000 (65%)\n",
            "\n",
            "\n",
            "Train Epoch 38: lr = 0.1\n",
            "Train Epoch: 38 [5000/50000 (10%)]\tTrain Loss: 0.982845\n",
            "Train Epoch: 38 [10000/50000 (20%)]\tTrain Loss: 0.894100\n",
            "Train Epoch: 38 [15000/50000 (30%)]\tTrain Loss: 0.940521\n",
            "Train Epoch: 38 [20000/50000 (40%)]\tTrain Loss: 0.919029\n",
            "Train Epoch: 38 [25000/50000 (50%)]\tTrain Loss: 0.885873\n",
            "Train Epoch: 38 [30000/50000 (60%)]\tTrain Loss: 0.950617\n",
            "Train Epoch: 38 [35000/50000 (70%)]\tTrain Loss: 0.927476\n",
            "Train Epoch: 38 [40000/50000 (80%)]\tTrain Loss: 0.947964\n",
            "Train Epoch: 38 [45000/50000 (90%)]\tTrain Loss: 0.954590\n",
            "\n",
            "Test set: Test loss: 1.0012, Accuracy: 3415/5000 (68%)\n",
            "\n",
            "\n",
            "Train Epoch 39: lr = 0.1\n",
            "Train Epoch: 39 [5000/50000 (10%)]\tTrain Loss: 0.889366\n",
            "Train Epoch: 39 [10000/50000 (20%)]\tTrain Loss: 0.962550\n",
            "Train Epoch: 39 [15000/50000 (30%)]\tTrain Loss: 0.937465\n",
            "Train Epoch: 39 [20000/50000 (40%)]\tTrain Loss: 0.947853\n",
            "Train Epoch: 39 [25000/50000 (50%)]\tTrain Loss: 0.927957\n",
            "Train Epoch: 39 [30000/50000 (60%)]\tTrain Loss: 0.906888\n",
            "Train Epoch: 39 [35000/50000 (70%)]\tTrain Loss: 0.889520\n",
            "Train Epoch: 39 [40000/50000 (80%)]\tTrain Loss: 0.994596\n",
            "Train Epoch: 39 [45000/50000 (90%)]\tTrain Loss: 0.959614\n",
            "\n",
            "Test set: Test loss: 1.1481, Accuracy: 3170/5000 (63%)\n",
            "\n",
            "\n",
            "Train Epoch 40: lr = 0.1\n",
            "Train Epoch: 40 [5000/50000 (10%)]\tTrain Loss: 1.061879\n",
            "Train Epoch: 40 [10000/50000 (20%)]\tTrain Loss: 0.970794\n",
            "Train Epoch: 40 [15000/50000 (30%)]\tTrain Loss: 0.945002\n",
            "Train Epoch: 40 [20000/50000 (40%)]\tTrain Loss: 1.029876\n",
            "Train Epoch: 40 [25000/50000 (50%)]\tTrain Loss: 1.110652\n",
            "Train Epoch: 40 [30000/50000 (60%)]\tTrain Loss: 0.973345\n",
            "Train Epoch: 40 [35000/50000 (70%)]\tTrain Loss: 0.940929\n",
            "Train Epoch: 40 [40000/50000 (80%)]\tTrain Loss: 0.952659\n",
            "Train Epoch: 40 [45000/50000 (90%)]\tTrain Loss: 0.950666\n",
            "\n",
            "Test set: Test loss: 0.9721, Accuracy: 3441/5000 (69%)\n",
            "\n",
            "Better accuracy at Epoch 40: accuracy = 68.82%\n",
            "\n",
            "Train Epoch 41: lr = 0.1\n",
            "Train Epoch: 41 [5000/50000 (10%)]\tTrain Loss: 0.910981\n",
            "Train Epoch: 41 [10000/50000 (20%)]\tTrain Loss: 1.000009\n",
            "Train Epoch: 41 [15000/50000 (30%)]\tTrain Loss: 0.990781\n",
            "Train Epoch: 41 [20000/50000 (40%)]\tTrain Loss: 0.973416\n",
            "Train Epoch: 41 [25000/50000 (50%)]\tTrain Loss: 0.923548\n",
            "Train Epoch: 41 [30000/50000 (60%)]\tTrain Loss: 0.953086\n",
            "Train Epoch: 41 [35000/50000 (70%)]\tTrain Loss: 0.897633\n",
            "Train Epoch: 41 [40000/50000 (80%)]\tTrain Loss: 0.898259\n",
            "Train Epoch: 41 [45000/50000 (90%)]\tTrain Loss: 0.890946\n",
            "\n",
            "Test set: Test loss: 1.0268, Accuracy: 3387/5000 (68%)\n",
            "\n",
            "\n",
            "Train Epoch 42: lr = 0.1\n",
            "Train Epoch: 42 [5000/50000 (10%)]\tTrain Loss: 0.890028\n",
            "Train Epoch: 42 [10000/50000 (20%)]\tTrain Loss: 0.910690\n",
            "Train Epoch: 42 [15000/50000 (30%)]\tTrain Loss: 0.894524\n",
            "Train Epoch: 42 [20000/50000 (40%)]\tTrain Loss: 0.966529\n",
            "Train Epoch: 42 [25000/50000 (50%)]\tTrain Loss: 0.935562\n",
            "Train Epoch: 42 [30000/50000 (60%)]\tTrain Loss: 0.963261\n",
            "Train Epoch: 42 [35000/50000 (70%)]\tTrain Loss: 0.959772\n",
            "Train Epoch: 42 [40000/50000 (80%)]\tTrain Loss: 0.968844\n",
            "Train Epoch: 42 [45000/50000 (90%)]\tTrain Loss: 0.980244\n",
            "\n",
            "Test set: Test loss: 1.0718, Accuracy: 3270/5000 (65%)\n",
            "\n",
            "\n",
            "Train Epoch 43: lr = 0.1\n",
            "Train Epoch: 43 [5000/50000 (10%)]\tTrain Loss: 1.004872\n",
            "Train Epoch: 43 [10000/50000 (20%)]\tTrain Loss: 0.889459\n",
            "Train Epoch: 43 [15000/50000 (30%)]\tTrain Loss: 0.889297\n",
            "Train Epoch: 43 [20000/50000 (40%)]\tTrain Loss: 0.880385\n",
            "Train Epoch: 43 [25000/50000 (50%)]\tTrain Loss: 1.008179\n",
            "Train Epoch: 43 [30000/50000 (60%)]\tTrain Loss: 0.959267\n",
            "Train Epoch: 43 [35000/50000 (70%)]\tTrain Loss: 0.951572\n",
            "Train Epoch: 43 [40000/50000 (80%)]\tTrain Loss: 0.864850\n",
            "Train Epoch: 43 [45000/50000 (90%)]\tTrain Loss: 0.909173\n",
            "\n",
            "Test set: Test loss: 0.9562, Accuracy: 3449/5000 (69%)\n",
            "\n",
            "Better accuracy at Epoch 43: accuracy = 68.98%\n",
            "Better loss at Epoch 43: loss = 0.9561693543195724%\n",
            "\n",
            "Train Epoch 44: lr = 0.1\n",
            "Train Epoch: 44 [5000/50000 (10%)]\tTrain Loss: 0.875874\n",
            "Train Epoch: 44 [10000/50000 (20%)]\tTrain Loss: 0.954500\n",
            "Train Epoch: 44 [15000/50000 (30%)]\tTrain Loss: 0.936015\n",
            "Train Epoch: 44 [20000/50000 (40%)]\tTrain Loss: 0.895852\n",
            "Train Epoch: 44 [25000/50000 (50%)]\tTrain Loss: 0.935822\n",
            "Train Epoch: 44 [30000/50000 (60%)]\tTrain Loss: 0.885621\n",
            "Train Epoch: 44 [35000/50000 (70%)]\tTrain Loss: 0.888445\n",
            "Train Epoch: 44 [40000/50000 (80%)]\tTrain Loss: 0.879370\n",
            "Train Epoch: 44 [45000/50000 (90%)]\tTrain Loss: 1.104435\n",
            "\n",
            "Test set: Test loss: 1.0553, Accuracy: 3295/5000 (66%)\n",
            "\n",
            "\n",
            "Train Epoch 45: lr = 0.1\n",
            "Train Epoch: 45 [5000/50000 (10%)]\tTrain Loss: 0.979204\n",
            "Train Epoch: 45 [10000/50000 (20%)]\tTrain Loss: 0.945336\n",
            "Train Epoch: 45 [15000/50000 (30%)]\tTrain Loss: 0.909616\n",
            "Train Epoch: 45 [20000/50000 (40%)]\tTrain Loss: 0.945136\n",
            "Train Epoch: 45 [25000/50000 (50%)]\tTrain Loss: 0.952466\n",
            "Train Epoch: 45 [30000/50000 (60%)]\tTrain Loss: 0.988138\n",
            "Train Epoch: 45 [35000/50000 (70%)]\tTrain Loss: 0.986206\n",
            "Train Epoch: 45 [40000/50000 (80%)]\tTrain Loss: 0.932167\n",
            "Train Epoch: 45 [45000/50000 (90%)]\tTrain Loss: 0.920932\n",
            "\n",
            "Test set: Test loss: 0.9846, Accuracy: 3354/5000 (67%)\n",
            "\n",
            "\n",
            "Train Epoch 46: lr = 0.1\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "error",
          "ename": "KeyboardInterrupt",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-11-9dd446c687eb>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mget_ipython\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrun_cell_magic\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'time'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m''\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'#training/testing\\nloss_list = []\\nloss_list_test = []\\niteration_list = []\\naccuracy_list = []\\n# execution time checking\\nexecution_time = []\\n\\nPATH = \\'/content/squeezenet_acc.pth\\'\\nbest_model_wts = copy.deepcopy(model.state_dict())\\nPATH2 = \\'/content/squeezenet_loss.pth\\'\\nbest_model_loss = copy.deepcopy(model.state_dict())\\n\\n#PATH = \\'/content/modified_mnist_effb2_pre_original_latest.pth\\'\\n#latest_model_wts = copy.deepcopy(model.state_dict())\\nbest_loss = float(\"inf\")\\nbest_acc = 0.0\\n#set number of epochs \\nnum_epochs = 100\\nfor epoch in range(num_epochs):\\n    #print(\"\\\\nTrain Epoch {}: lr = {}\".format(epoch, exp_lr_scheduler.get_lr()[0]))\\n    print(\"\\\\nTrain Epoch {}: lr = {}\".format(epoch,learning_rate))\\n\\n    start = timeit.default_timer()\\n    loss_list.append(train(epoch, train_loader, model, error, optimizer, batch_size))\\n    loss_test, accuracy = test(test_loader, model, error, batch_size)\\n    stop = timeit.default_timer()\\n    etime = stop - start\\n    execution_time.append(etime)\\n\\n    loss_list_test.append(loss_test)\\n    accuracy_list.append(accuracy)\\n    iteration_list.append(epoch)\\n    #exp_lr_scheduler.step()\\n\\n    #saving model with best acc \\n    if accuracy > best_acc:\\n      print(f\"Better accuracy at Epoch {epoch}: accuracy = {accuracy}%\")\\n      best_acc = accur...\n\u001b[0m",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/IPython/core/interactiveshell.py\u001b[0m in \u001b[0;36mrun_cell_magic\u001b[0;34m(self, magic_name, line, cell)\u001b[0m\n\u001b[1;32m   2115\u001b[0m             \u001b[0mmagic_arg_s\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvar_expand\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mline\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mstack_depth\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2116\u001b[0m             \u001b[0;32mwith\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbuiltin_trap\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2117\u001b[0;31m                 \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmagic_arg_s\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcell\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   2118\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0mresult\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2119\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m</usr/local/lib/python3.6/dist-packages/decorator.py:decorator-gen-60>\u001b[0m in \u001b[0;36mtime\u001b[0;34m(self, line, cell, local_ns)\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/IPython/core/magic.py\u001b[0m in \u001b[0;36m<lambda>\u001b[0;34m(f, *a, **k)\u001b[0m\n\u001b[1;32m    186\u001b[0m     \u001b[0;31m# but it's overkill for just that one bit of state.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    187\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mmagic_deco\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0marg\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 188\u001b[0;31m         \u001b[0mcall\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mlambda\u001b[0m \u001b[0mf\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0ma\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mk\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mf\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0ma\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mk\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    189\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    190\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mcallable\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0marg\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/IPython/core/magics/execution.py\u001b[0m in \u001b[0;36mtime\u001b[0;34m(self, line, cell, local_ns)\u001b[0m\n\u001b[1;32m   1191\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1192\u001b[0m             \u001b[0mst\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mclock2\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1193\u001b[0;31m             \u001b[0mexec\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcode\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mglob\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlocal_ns\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1194\u001b[0m             \u001b[0mend\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mclock2\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1195\u001b[0m             \u001b[0mout\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m<timed exec>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n",
            "\u001b[0;32m<ipython-input-4-c3273aca21cc>\u001b[0m in \u001b[0;36mtrain\u001b[0;34m(epoch, train_loader, model, error, optimizer, batch_size)\u001b[0m\n\u001b[1;32m     18\u001b[0m         \u001b[0moptimizer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mzero_grad\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     19\u001b[0m         \u001b[0;31m# Forward propagation\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 20\u001b[0;31m         \u001b[0moutputs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtrain\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     21\u001b[0m         \u001b[0;31m# Calculate softmax and cross entropy loss\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     22\u001b[0m         \u001b[0mloss\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0merror\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0moutputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlabels\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, *input, **kwargs)\u001b[0m\n\u001b[1;32m    539\u001b[0m             \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_slow_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    540\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 541\u001b[0;31m             \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    542\u001b[0m         \u001b[0;32mfor\u001b[0m \u001b[0mhook\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_forward_hooks\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    543\u001b[0m             \u001b[0mhook_result\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mhook\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mresult\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m<ipython-input-8-f78ac235f9f0>\u001b[0m in \u001b[0;36mforward\u001b[0;34m(self, x)\u001b[0m\n\u001b[1;32m    101\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    102\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 103\u001b[0;31m         \u001b[0mx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfeatures\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    104\u001b[0m         \u001b[0mx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mclassifier\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    105\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mflatten\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, *input, **kwargs)\u001b[0m\n\u001b[1;32m    539\u001b[0m             \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_slow_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    540\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 541\u001b[0;31m             \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    542\u001b[0m         \u001b[0;32mfor\u001b[0m \u001b[0mhook\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_forward_hooks\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    543\u001b[0m             \u001b[0mhook_result\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mhook\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mresult\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/torch/nn/modules/container.py\u001b[0m in \u001b[0;36mforward\u001b[0;34m(self, input)\u001b[0m\n\u001b[1;32m     90\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     91\u001b[0m         \u001b[0;32mfor\u001b[0m \u001b[0mmodule\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_modules\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 92\u001b[0;31m             \u001b[0minput\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmodule\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     93\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     94\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, *input, **kwargs)\u001b[0m\n\u001b[1;32m    539\u001b[0m             \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_slow_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    540\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 541\u001b[0;31m             \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    542\u001b[0m         \u001b[0;32mfor\u001b[0m \u001b[0mhook\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_forward_hooks\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    543\u001b[0m             \u001b[0mhook_result\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mhook\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mresult\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m<ipython-input-8-f78ac235f9f0>\u001b[0m in \u001b[0;36mforward\u001b[0;34m(self, x)\u001b[0m\n\u001b[1;32m     35\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mexpand1x1_activation\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mexpand1x1\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     36\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mexpand3x3_activation\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mexpand3x3\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 37\u001b[0;31m         ], 1)\n\u001b[0m\u001b[1;32m     38\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     39\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DEpPEK4ZZbPo",
        "colab_type": "code",
        "outputId": "67f99fcf-da38-4726-f943-315b7962fb84",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 608
        }
      },
      "source": [
        "  # visualization loss \n",
        "plt.plot(iteration_list,loss_list, label='Training loss')\n",
        "plt.plot(iteration_list,loss_list_test, label='Validation loss')\n",
        "plt.xlabel(\"Number of iteration\")\n",
        "plt.ylabel(\"Loss\")\n",
        "plt.title(\"Squeezenet (Original)\")\n",
        "plt.legend()\n",
        "plt.show()\n",
        "\n",
        "# visualization accuracy \n",
        "plt.plot(iteration_list,accuracy_list,color = \"red\")\n",
        "plt.xlabel(\"Number of iteration\")\n",
        "plt.ylabel(\"Accuracy\")\n",
        "plt.title(\"Squeezenet (Original)\")\n",
        "plt.show()\n",
        "\n",
        "print(f\"Mean time per epoch {np.mean(execution_time)} seconds\") # returns time in sec \n",
        "print(f\"Best accuracy: {best_acc}  Best training loss: {min(loss_list)}  Best validation loss: {min(loss_list_test)}\")\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEWCAYAAABrDZDcAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\niR8FEBD783IAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX4AAAEWCAYAAABhffzLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Mean time per epoch 24.28885762584782 seconds\n",
            "Best accuracy: 68.98  Best training loss: 0.6064860820770264  Best validation loss: 0.9561693543195724\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jhTWsbg_ejfb",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "print(str(iteration_list))\n",
        "print(str(loss_list))\n",
        "print(str(loss_list_test))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "tAoiZY-gnjc1",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5XUt6oJNnkHO",
        "colab_type": "text"
      },
      "source": [
        "### testing adding layers"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ogEv-s47np-z",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.nn.init as init\n",
        "import torch.utils.model_zoo as model_zoo\n",
        "\n",
        "__all__ = ['SqueezeNet', 'squeezenet1_0', 'squeezenet1_1']\n",
        "\n",
        "model_urls = {\n",
        "    'squeezenet1_0': 'https://download.pytorch.org/models/squeezenet1_0-a815701f.pth',\n",
        "    'squeezenet1_1': 'https://download.pytorch.org/models/squeezenet1_1-f364aa15.pth',\n",
        "}\n",
        "\n",
        "\n",
        "class Fire(nn.Module):\n",
        "\n",
        "    def __init__(self, inplanes, squeeze_planes,\n",
        "                 expand1x1_planes, expand3x3_planes):\n",
        "        super(Fire, self).__init__()\n",
        "        self.inplanes = inplanes\n",
        "        self.squeeze = nn.Conv2d(inplanes, squeeze_planes, kernel_size=1)\n",
        "        #self.squeeze_activation = nn.ReLU(inplace=True)\n",
        "        self.squeeze_activation = nn.Tanh(inplace=True)\n",
        "        self.expand1x1 = nn.Conv2d(squeeze_planes, expand1x1_planes,\n",
        "                                   kernel_size=1)\n",
        "        #self.expand1x1_activation = nn.ReLU(inplace=True)\n",
        "        self.expand1x1_activation = nn.Tanh(inplace=True)\n",
        "        self.expand3x3 = nn.Conv2d(squeeze_planes, expand3x3_planes,\n",
        "                                   kernel_size=3, padding=1)\n",
        "        #self.expand3x3_activation = nn.ReLU(inplace=True)\n",
        "        self.expand3x3_activation = nn.Tanh(inplace=True)\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = self.squeeze_activation(self.squeeze(x))\n",
        "        return torch.cat([\n",
        "            self.expand1x1_activation(self.expand1x1(x)),\n",
        "            self.expand3x3_activation(self.expand3x3(x))\n",
        "        ], 1)\n",
        "\n",
        "\n",
        "class SqueezeNet(nn.Module):\n",
        "\n",
        "    def __init__(self, version='1_0', num_classes=1000):\n",
        "        super(SqueezeNet, self).__init__()\n",
        "        self.num_classes = num_classes\n",
        "        if version == '1_0':\n",
        "            self.features = nn.Sequential(\n",
        "                nn.Conv2d(3, 96, kernel_size=7, stride=2),\n",
        "                nn.ReLU(inplace=True),\n",
        "                nn.MaxPool2d(kernel_size=3, stride=2, ceil_mode=True),\n",
        "                Fire(96, 16, 64, 64),\n",
        "                Fire(128, 16, 64, 64),\n",
        "                Fire(128, 32, 128, 128),\n",
        "                nn.MaxPool2d(kernel_size=3, stride=2, ceil_mode=True),\n",
        "                Fire(256, 32, 128, 128),\n",
        "                Fire(256, 48, 192, 192),\n",
        "                Fire(384, 48, 192, 192),\n",
        "                Fire(384, 64, 256, 256),\n",
        "                nn.MaxPool2d(kernel_size=3, stride=2, ceil_mode=True),\n",
        "                Fire(512, 64, 256, 256),\n",
        "            )\n",
        "        elif version == '1_1':\n",
        "            self.features = nn.Sequential(\n",
        "                nn.Conv2d(3, 64, kernel_size=3, stride=2),\n",
        "                nn.ReLU(inplace=True),\n",
        "                nn.MaxPool2d(kernel_size=3, stride=2, ceil_mode=True),\n",
        "                Fire(64, 16, 64, 64),\n",
        "                Fire(128, 16, 64, 64),\n",
        "                nn.MaxPool2d(kernel_size=3, stride=2, ceil_mode=True),\n",
        "                Fire(128, 32, 128, 128),\n",
        "                Fire(256, 32, 128, 128),\n",
        "                nn.MaxPool2d(kernel_size=3, stride=2, ceil_mode=True),\n",
        "                Fire(256, 48, 192, 192),\n",
        "                Fire(384, 48, 192, 192),\n",
        "                Fire(384, 64, 256, 256),\n",
        "                Fire(512, 64, 256, 256),\n",
        "            )\n",
        "        else:\n",
        "            # FIXME: Is this needed? SqueezeNet should only be called from the\n",
        "            # FIXME: squeezenet1_x() functions\n",
        "            # FIXME: This checking is not done for the other models\n",
        "            raise ValueError(\"Unsupported SqueezeNet version {version}:\"\n",
        "                             \"1_0 or 1_1 expected\".format(version=version))\n",
        "\n",
        "        # Final convolution is initialized differently from the rest\n",
        "        final_conv = nn.Conv2d(512, self.num_classes, kernel_size=1)\n",
        "        self.classifier = nn.Sequential(\n",
        "            nn.Dropout(p=0.5),\n",
        "            final_conv,\n",
        "            nn.ReLU(inplace=True),\n",
        "            nn.AdaptiveAvgPool2d((1, 1))\n",
        "        )\n",
        "\n",
        "        for m in self.modules():\n",
        "            if isinstance(m, nn.Conv2d):\n",
        "                if m is final_conv:\n",
        "                    init.normal_(m.weight, mean=0.0, std=0.01)\n",
        "                else:\n",
        "                    init.kaiming_uniform_(m.weight)\n",
        "                if m.bias is not None:\n",
        "                    init.constant_(m.bias, 0)\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = self.features(x)\n",
        "        x = self.classifier(x)\n",
        "        return torch.flatten(x, 1)\n",
        "\n",
        "\n",
        "def _squeezenet(version, pretrained, progress, **kwargs):\n",
        "    model = SqueezeNet(version, **kwargs)\n",
        "    if pretrained:\n",
        "        arch = 'squeezenet' + version\n",
        "        state_dict = load_state_dict_from_url(model_urls[arch],\n",
        "                                              progress=progress)\n",
        "        model.load_state_dict(state_dict)\n",
        "    return model\n",
        "\n",
        "\n",
        "def squeezenet1_0(pretrained=False, progress=True, **kwargs):\n",
        "    r\"\"\"SqueezeNet model architecture from the `\"SqueezeNet: AlexNet-level\n",
        "    accuracy with 50x fewer parameters and <0.5MB model size\"\n",
        "    <https://arxiv.org/abs/1602.07360>`_ paper.\n",
        "    Args:\n",
        "        pretrained (bool): If True, returns a model pre-trained on ImageNet\n",
        "        progress (bool): If True, displays a progress bar of the download to stderr\n",
        "    \"\"\"\n",
        "    return _squeezenet('1_0', pretrained, progress, **kwargs)\n",
        "\n",
        "\n",
        "def squeezenet1_1(pretrained=False, progress=True, **kwargs):\n",
        "    r\"\"\"SqueezeNet 1.1 model from the `official SqueezeNet repo\n",
        "    <https://github.com/DeepScale/SqueezeNet/tree/master/SqueezeNet_v1.1>`_.\n",
        "    SqueezeNet 1.1 has 2.4x less computation and slightly fewer parameters\n",
        "    than SqueezeNet 1.0, without sacrificing accuracy.\n",
        "    Args:\n",
        "        pretrained (bool): If True, returns a model pre-trained on ImageNet\n",
        "        progress (bool): If True, displays a progress bar of the download to stderr\n",
        "    \"\"\"\n",
        "    return _squeezenet('1_1', pretrained, progress, **kwargs)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "SOF3GmFBfZQM",
        "colab_type": "text"
      },
      "source": [
        "## squeezing 1.1"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kbrIoT93f68T",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "\n",
        "#change the line below for different models\n",
        "\n",
        "# import torchvision.models as models\n",
        "# model = models.squeezenet1_0(pretrained=True)\n",
        "\n",
        "model = squeezenet1_1(num_classes=10)\n",
        "\n",
        "# print(model)\n",
        "\n",
        "model = model.to(device=device, dtype=torch.float)\n",
        "\n",
        "# Cross Entropy Loss \n",
        "error = CrossEntropyLoss().to(device=device, dtype=torch.float)\n",
        "\n",
        "#Optimizer\n",
        "learning_rate = 0.1\n",
        "optimizer = adabound.AdaBound(model.parameters(), lr=1e-3, final_lr=learning_rate)\n",
        "\n",
        "#Optimizer adam\n",
        "# learning_rate = 0.04\n",
        "# optimizer = Adam(model.parameters(), lr=learning_rate)\n",
        "# optimizer = SGD(model.parameters(), lr=learning_rate, momentum=0.9, dampening=0, weight_decay=0.0002, nesterov=False)\n",
        "# optimizer = SGD(model.parameters(), lr=learning_rate)\n",
        "\n",
        "#adaptive learning rate\n",
        "# how many epoc before decreasing learning rate\n",
        "#exp_lr_scheduler = lr_scheduler.StepLR(optimizer, step_size=5, gamma=0.1)\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jNsLpPjHfYH_",
        "colab_type": "code",
        "outputId": "0583d56f-088c-43bc-e3f5-02d12ad4dffe",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 510
        }
      },
      "source": [
        "%%time\n",
        "#training/testing\n",
        "loss_list = []\n",
        "loss_list_test = []\n",
        "iteration_list = []\n",
        "accuracy_list = []\n",
        "# execution time checking\n",
        "execution_time = []\n",
        "\n",
        "PATH = '/content/squeezenet_acc.pth'\n",
        "best_model_wts = copy.deepcopy(model.state_dict())\n",
        "PATH2 = '/content/squeezenet_loss.pth'\n",
        "best_model_loss = copy.deepcopy(model.state_dict())\n",
        "\n",
        "#PATH = '/content/modified_mnist_effb2_pre_original_latest.pth'\n",
        "#latest_model_wts = copy.deepcopy(model.state_dict())\n",
        "best_loss = float(\"inf\")\n",
        "best_acc = 0.0\n",
        "#set number of epochs \n",
        "num_epochs = 25\n",
        "for epoch in range(num_epochs):\n",
        "    #print(\"\\nTrain Epoch {}: lr = {}\".format(epoch, exp_lr_scheduler.get_lr()[0]))\n",
        "    print(\"\\nTrain Epoch {}: lr = {}\".format(epoch,learning_rate))\n",
        "\n",
        "    start = timeit.default_timer()\n",
        "    loss_list.append(train(epoch, train_loader, model, error, optimizer, batch_size))\n",
        "    loss_test, accuracy = test(test_loader, model, error, batch_size)\n",
        "    stop = timeit.default_timer()\n",
        "    etime = stop - start\n",
        "    execution_time.append(etime)\n",
        "\n",
        "    loss_list_test.append(loss_test)\n",
        "    accuracy_list.append(accuracy)\n",
        "    iteration_list.append(epoch)\n",
        "    #exp_lr_scheduler.step()\n",
        "\n",
        "    #saving model with best acc \n",
        "    if accuracy > best_acc:\n",
        "      print(f\"Better accuracy at Epoch {epoch}: accuracy = {accuracy}%\")\n",
        "      best_acc = accuracy\n",
        "      best_model_wts = copy.deepcopy(model.state_dict())\n",
        "      torch.save(best_model_wts, PATH)\n",
        "    #saving model with best loss\n",
        "    if loss_test < best_loss:\n",
        "      print(f\"Better loss at Epoch {epoch}: loss = {loss_test}%\")\n",
        "      best_loss = loss_test\n",
        "      best_model_loss = copy.deepcopy(model.state_dict())\n",
        "      torch.save(best_model_loss, PATH2)\n",
        "#load model\n",
        "# model = Net()\n",
        "# model.load_state_dict(torch.load(PATH))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\n",
            "Train Epoch 0: lr = 0.1\n",
            "Train Epoch: 0 [5000/50000 (10%)]\tTrain Loss: 2.325610\n",
            "Train Epoch: 0 [10000/50000 (20%)]\tTrain Loss: 2.302584\n",
            "Train Epoch: 0 [15000/50000 (30%)]\tTrain Loss: 2.302584\n",
            "Train Epoch: 0 [20000/50000 (40%)]\tTrain Loss: 2.302584\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "error",
          "ename": "KeyboardInterrupt",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-33-0ae23a9cf250>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mget_ipython\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrun_cell_magic\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'time'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m''\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'#training/testing\\nloss_list = []\\nloss_list_test = []\\niteration_list = []\\naccuracy_list = []\\n# execution time checking\\nexecution_time = []\\n\\nPATH = \\'/content/squeezenet_acc.pth\\'\\nbest_model_wts = copy.deepcopy(model.state_dict())\\nPATH2 = \\'/content/squeezenet_loss.pth\\'\\nbest_model_loss = copy.deepcopy(model.state_dict())\\n\\n#PATH = \\'/content/modified_mnist_effb2_pre_original_latest.pth\\'\\n#latest_model_wts = copy.deepcopy(model.state_dict())\\nbest_loss = float(\"inf\")\\nbest_acc = 0.0\\n#set number of epochs \\nnum_epochs = 25\\nfor epoch in range(num_epochs):\\n    #print(\"\\\\nTrain Epoch {}: lr = {}\".format(epoch, exp_lr_scheduler.get_lr()[0]))\\n    print(\"\\\\nTrain Epoch {}: lr = {}\".format(epoch,learning_rate))\\n\\n    start = timeit.default_timer()\\n    loss_list.append(train(epoch, train_loader, model, error, optimizer, batch_size))\\n    loss_test, accuracy = test(test_loader, model, error, batch_size)\\n    stop = timeit.default_timer()\\n    etime = stop - start\\n    execution_time.append(etime)\\n\\n    loss_list_test.append(loss_test)\\n    accuracy_list.append(accuracy)\\n    iteration_list.append(epoch)\\n    #exp_lr_scheduler.step()\\n\\n    #saving model with best acc \\n    if accuracy > best_acc:\\n      print(f\"Better accuracy at Epoch {epoch}: accuracy = {accuracy}%\")\\n      best_acc = accura...\n\u001b[0m",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/IPython/core/interactiveshell.py\u001b[0m in \u001b[0;36mrun_cell_magic\u001b[0;34m(self, magic_name, line, cell)\u001b[0m\n\u001b[1;32m   2115\u001b[0m             \u001b[0mmagic_arg_s\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvar_expand\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mline\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mstack_depth\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2116\u001b[0m             \u001b[0;32mwith\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbuiltin_trap\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2117\u001b[0;31m                 \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmagic_arg_s\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcell\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   2118\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0mresult\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2119\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m</usr/local/lib/python3.6/dist-packages/decorator.py:decorator-gen-60>\u001b[0m in \u001b[0;36mtime\u001b[0;34m(self, line, cell, local_ns)\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/IPython/core/magic.py\u001b[0m in \u001b[0;36m<lambda>\u001b[0;34m(f, *a, **k)\u001b[0m\n\u001b[1;32m    186\u001b[0m     \u001b[0;31m# but it's overkill for just that one bit of state.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    187\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mmagic_deco\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0marg\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 188\u001b[0;31m         \u001b[0mcall\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mlambda\u001b[0m \u001b[0mf\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0ma\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mk\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mf\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0ma\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mk\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    189\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    190\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mcallable\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0marg\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/IPython/core/magics/execution.py\u001b[0m in \u001b[0;36mtime\u001b[0;34m(self, line, cell, local_ns)\u001b[0m\n\u001b[1;32m   1191\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1192\u001b[0m             \u001b[0mst\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mclock2\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1193\u001b[0;31m             \u001b[0mexec\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcode\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mglob\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlocal_ns\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1194\u001b[0m             \u001b[0mend\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mclock2\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1195\u001b[0m             \u001b[0mout\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m<timed exec>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n",
            "\u001b[0;32m<ipython-input-4-c3273aca21cc>\u001b[0m in \u001b[0;36mtrain\u001b[0;34m(epoch, train_loader, model, error, optimizer, batch_size)\u001b[0m\n\u001b[1;32m     25\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     26\u001b[0m         \u001b[0;31m# Calculating gradients\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 27\u001b[0;31m         \u001b[0mloss\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbackward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     28\u001b[0m         \u001b[0;31m# Update parameters\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     29\u001b[0m         \u001b[0moptimizer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mstep\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/torch/tensor.py\u001b[0m in \u001b[0;36mbackward\u001b[0;34m(self, gradient, retain_graph, create_graph)\u001b[0m\n\u001b[1;32m    164\u001b[0m                 \u001b[0mproducts\u001b[0m\u001b[0;34m.\u001b[0m \u001b[0mDefaults\u001b[0m \u001b[0mto\u001b[0m\u001b[0;31m \u001b[0m\u001b[0;31m`\u001b[0m\u001b[0;31m`\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;31m`\u001b[0m\u001b[0;31m`\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    165\u001b[0m         \"\"\"\n\u001b[0;32m--> 166\u001b[0;31m         \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mautograd\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbackward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgradient\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mretain_graph\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcreate_graph\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    167\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    168\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mregister_hook\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mhook\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/torch/autograd/__init__.py\u001b[0m in \u001b[0;36mbackward\u001b[0;34m(tensors, grad_tensors, retain_graph, create_graph, grad_variables)\u001b[0m\n\u001b[1;32m     97\u001b[0m     Variable._execution_engine.run_backward(\n\u001b[1;32m     98\u001b[0m         \u001b[0mtensors\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgrad_tensors\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mretain_graph\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcreate_graph\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 99\u001b[0;31m         allow_unreachable=True)  # allow_unreachable flag\n\u001b[0m\u001b[1;32m    100\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    101\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2tQdRvg0gBuF",
        "colab_type": "code",
        "outputId": "4ee44259-6064-4461-b784-53e15b7b94b4",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 607
        }
      },
      "source": [
        "  # visualization loss \n",
        "plt.plot(iteration_list,loss_list, label='Training loss')\n",
        "plt.plot(iteration_list,loss_list_test, label='Validation loss')\n",
        "plt.xlabel(\"Number of iteration\")\n",
        "plt.ylabel(\"Loss\")\n",
        "plt.title(\"Squeezenet 1.1 (Original)\")\n",
        "plt.legend()\n",
        "plt.show()\n",
        "\n",
        "# visualization accuracy \n",
        "plt.plot(iteration_list,accuracy_list,color = \"red\")\n",
        "plt.xlabel(\"Number of iteration\")\n",
        "plt.ylabel(\"Accuracy\")\n",
        "plt.title(\"Squeezenet 1.1 (Original)\")\n",
        "plt.show()\n",
        "\n",
        "print(f\"Mean time per epoch {np.mean(execution_time)} seconds\") # returns time in sec \n",
        "print(f\"Best accuracy: {best_acc}  Best training loss: {min(loss_list)}  Best validation loss: {min(loss_list_test)}\")\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYYAAAEWCAYAAABi5jCmAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nwczM8jgxmJlZnv8HMh8z9+i3d8kAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEWCAYAAABrDZDcAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nFnsTk+46YVgAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Mean time per epoch 24.102241231279994 seconds\n",
            "Best accuracy: 56.0  Best training loss: 0.915762186050415  Best validation loss: 1.3321439516544342\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OYCYDYWOgFoR",
        "colab_type": "code",
        "outputId": "7887d49e-b7d5-4751-c86f-e3da6cb57ee9",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 88
        }
      },
      "source": [
        "print(str(iteration_list))\n",
        "print(str(loss_list))\n",
        "print(str(loss_list_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24]\n",
            "[2.1838796138763428, 2.0178678035736084, 2.0018107891082764, 1.9213051795959473, 1.9051405191421509, 2.060081958770752, 1.9278173446655273, 1.7947484254837036, 1.6226844787597656, 1.6778972148895264, 1.66666841506958, 1.2055578231811523, 1.095021367073059, 1.1353867053985596, 1.5495797395706177, 1.1674708127975464, 1.4456303119659424, 1.2469276189804077, 1.2465100288391113, 1.155672550201416, 1.4057934284210205, 1.2752779722213745, 1.2141385078430176, 0.915762186050415, 0.9800704121589661]\n",
            "[2.121135513782501, 2.118862378001213, 2.0969004899263384, 2.013542518019676, 1.9705626732110977, 1.9798394191265105, 1.845936484336853, 1.675332503914833, 1.5929853576421737, 1.5656340289115906, 1.522431623339653, 1.5168981689214707, 1.4533483976125716, 1.4200146025419236, 1.3806409525871277, 1.4513233157992362, 1.391934596300125, 1.3997139877080917, 1.3558638465404511, 1.3708526888489723, 1.3737952956557273, 1.365614928007126, 1.3628060725331306, 1.399139752984047, 1.3321439516544342]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ik8CPhgtgJHz",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "O0CSnPUgHgwG",
        "colab_type": "code",
        "outputId": "6127367f-0877-4383-8e47-3fb958abfea3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 470
        }
      },
      "source": [
        "import matplotlib.pyplot as plt; plt.rcdefaults()\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "objects = ('airplane','bird', 'cat', 'deer',\t'dog', 'frog', 'horse', 'ship',\t'truck   ','    automobile')\n",
        "y_pos = np.arange(len(objects))\n",
        "y = [1000, 2000, 3000,4000,5000, 6000,7000, 8000]\n",
        "performance = [6000,6000,6000,6000,6000,6000,6000,6000,6000, 6000]\n",
        "\n",
        "plt.bar(y_pos, performance, width=0.5)\n",
        "plt.xticks(y_pos, objects)\n",
        "plt.yticks(y)\n",
        "plt.ylabel('Number of Samples From Particular Class')\n",
        "plt.title('Class Distribution in CIFAR-10 Dataset')\n",
        "from matplotlib.pyplot import figure\n",
        "figure(num=None, figsize=(15,15), dpi=80, facecolor='w', edgecolor='k')\n",
        "plt.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAlEAAAGzCAYAAAAPGELKAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1200x1200 with 0 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "99_on1FAHhhb",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}